2011-11-02 Richard Guenther <rguenther@suse.de>

[pf3gnuchains/gcc-fork.git] / gcc / tree-vect-stmts.c

/* Statement Analysis and Transformation for Vectorization
   Copyright (C) 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
   Free Software Foundation, Inc.
   Contributed by Dorit Naishlos <dorit@il.ibm.com>
   and Ira Rosen <irar@il.ibm.com>

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 "tm.h"
#include "ggc.h"
#include "tree.h"
#include "target.h"
#include "basic-block.h"
#include "tree-pretty-print.h"
#include "gimple-pretty-print.h"
#include "tree-flow.h"
#include "tree-dump.h"
#include "cfgloop.h"
#include "cfglayout.h"
#include "expr.h"
#include "recog.h"
#include "optabs.h"
#include "diagnostic-core.h"
#include "tree-vectorizer.h"
#include "langhooks.h"


/* Return a variable of type ELEM_TYPE[NELEMS].  */

static tree
create_vector_array (tree elem_type, unsigned HOST_WIDE_INT nelems)
{
  return create_tmp_var (build_array_type_nelts (elem_type, nelems),
                         "vect_array");
}

/* ARRAY is an array of vectors created by create_vector_array.
   Return an SSA_NAME for the vector in index N.  The reference
   is part of the vectorization of STMT and the vector is associated
   with scalar destination SCALAR_DEST.  */

static tree
read_vector_array (gimple stmt, gimple_stmt_iterator *gsi, tree scalar_dest,
                   tree array, unsigned HOST_WIDE_INT n)
{
  tree vect_type, vect, vect_name, array_ref;
  gimple new_stmt;

  gcc_assert (TREE_CODE (TREE_TYPE (array)) == ARRAY_TYPE);
  vect_type = TREE_TYPE (TREE_TYPE (array));
  vect = vect_create_destination_var (scalar_dest, vect_type);
  array_ref = build4 (ARRAY_REF, vect_type, array,
                      build_int_cst (size_type_node, n),
                      NULL_TREE, NULL_TREE);

  new_stmt = gimple_build_assign (vect, array_ref);
  vect_name = make_ssa_name (vect, new_stmt);
  gimple_assign_set_lhs (new_stmt, vect_name);
  vect_finish_stmt_generation (stmt, new_stmt, gsi);
  mark_symbols_for_renaming (new_stmt);

  return vect_name;
}

/* ARRAY is an array of vectors created by create_vector_array.
   Emit code to store SSA_NAME VECT in index N of the array.
   The store is part of the vectorization of STMT.  */

static void
write_vector_array (gimple stmt, gimple_stmt_iterator *gsi, tree vect,
                    tree array, unsigned HOST_WIDE_INT n)
{
  tree array_ref;
  gimple new_stmt;

  array_ref = build4 (ARRAY_REF, TREE_TYPE (vect), array,
                      build_int_cst (size_type_node, n),
                      NULL_TREE, NULL_TREE);

  new_stmt = gimple_build_assign (array_ref, vect);
  vect_finish_stmt_generation (stmt, new_stmt, gsi);
  mark_symbols_for_renaming (new_stmt);
}

/* PTR is a pointer to an array of type TYPE.  Return a representation
   of *PTR.  The memory reference replaces those in FIRST_DR
   (and its group).  */

static tree
create_array_ref (tree type, tree ptr, struct data_reference *first_dr)
{
  struct ptr_info_def *pi;
  tree mem_ref, alias_ptr_type;

  alias_ptr_type = reference_alias_ptr_type (DR_REF (first_dr));
  mem_ref = build2 (MEM_REF, type, ptr, build_int_cst (alias_ptr_type, 0));
  /* Arrays have the same alignment as their type.  */
  pi = get_ptr_info (ptr);
  pi->align = TYPE_ALIGN_UNIT (type);
  pi->misalign = 0;
  return mem_ref;
}

/* Utility functions used by vect_mark_stmts_to_be_vectorized.  */

/* Function vect_mark_relevant.

   Mark STMT as "relevant for vectorization" and add it to WORKLIST.  */

static void
vect_mark_relevant (VEC(gimple,heap) **worklist, gimple stmt,
                    enum vect_relevant relevant, bool live_p,
                    bool used_in_pattern)
{
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
  enum vect_relevant save_relevant = STMT_VINFO_RELEVANT (stmt_info);
  bool save_live_p = STMT_VINFO_LIVE_P (stmt_info);
  gimple pattern_stmt;

  if (vect_print_dump_info (REPORT_DETAILS))
    fprintf (vect_dump, "mark relevant %d, live %d.", relevant, live_p);

  /* If this stmt is an original stmt in a pattern, we might need to mark its
     related pattern stmt instead of the original stmt.  However, such stmts
     may have their own uses that are not in any pattern, in such cases the
     stmt itself should be marked.  */
  if (STMT_VINFO_IN_PATTERN_P (stmt_info))
    {
      bool found = false;
      if (!used_in_pattern)
        {
          imm_use_iterator imm_iter;
          use_operand_p use_p;
          gimple use_stmt;
          tree lhs;

          if (is_gimple_assign (stmt))
            lhs = gimple_assign_lhs (stmt);
          else
            lhs = gimple_call_lhs (stmt);

          /* This use is out of pattern use, if LHS has other uses that are
             pattern uses, we should mark the stmt itself, and not the pattern
             stmt.  */
          if (TREE_CODE (lhs) == SSA_NAME)
            FOR_EACH_IMM_USE_FAST (use_p, imm_iter, lhs)
              {
                if (is_gimple_debug (USE_STMT (use_p)))
                  continue;
                use_stmt = USE_STMT (use_p);

                if (vinfo_for_stmt (use_stmt)
                    && STMT_VINFO_IN_PATTERN_P (vinfo_for_stmt (use_stmt)))
                  {
                    found = true;
                    break;
                  }
              }
        }

      if (!found)
        {
          /* This is the last stmt in a sequence that was detected as a
             pattern that can potentially be vectorized.  Don't mark the stmt
             as relevant/live because it's not going to be vectorized.
             Instead mark the pattern-stmt that replaces it.  */

          pattern_stmt = STMT_VINFO_RELATED_STMT (stmt_info);

          if (vect_print_dump_info (REPORT_DETAILS))
            fprintf (vect_dump, "last stmt in pattern. don't mark"
                                " relevant/live.");
          stmt_info = vinfo_for_stmt (pattern_stmt);
          gcc_assert (STMT_VINFO_RELATED_STMT (stmt_info) == stmt);
          save_relevant = STMT_VINFO_RELEVANT (stmt_info);
          save_live_p = STMT_VINFO_LIVE_P (stmt_info);
          stmt = pattern_stmt;
        }
    }

  STMT_VINFO_LIVE_P (stmt_info) |= live_p;
  if (relevant > STMT_VINFO_RELEVANT (stmt_info))
    STMT_VINFO_RELEVANT (stmt_info) = relevant;

  if (STMT_VINFO_RELEVANT (stmt_info) == save_relevant
      && STMT_VINFO_LIVE_P (stmt_info) == save_live_p)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "already marked relevant/live.");
      return;
    }

  VEC_safe_push (gimple, heap, *worklist, stmt);
}


/* Function vect_stmt_relevant_p.

   Return true if STMT in loop that is represented by LOOP_VINFO is
   "relevant for vectorization".

   A stmt is considered "relevant for vectorization" if:
   - it has uses outside the loop.
   - it has vdefs (it alters memory).
   - control stmts in the loop (except for the exit condition).

   CHECKME: what other side effects would the vectorizer allow?  */

static bool
vect_stmt_relevant_p (gimple stmt, loop_vec_info loop_vinfo,
                      enum vect_relevant *relevant, bool *live_p)
{
  struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
  ssa_op_iter op_iter;
  imm_use_iterator imm_iter;
  use_operand_p use_p;
  def_operand_p def_p;

  *relevant = vect_unused_in_scope;
  *live_p = false;

  /* cond stmt other than loop exit cond.  */
  if (is_ctrl_stmt (stmt)
      && STMT_VINFO_TYPE (vinfo_for_stmt (stmt))
         != loop_exit_ctrl_vec_info_type)
    *relevant = vect_used_in_scope;

  /* changing memory.  */
  if (gimple_code (stmt) != GIMPLE_PHI)
    if (gimple_vdef (stmt))
      {
        if (vect_print_dump_info (REPORT_DETAILS))
          fprintf (vect_dump, "vec_stmt_relevant_p: stmt has vdefs.");
        *relevant = vect_used_in_scope;
      }

  /* uses outside the loop.  */
  FOR_EACH_PHI_OR_STMT_DEF (def_p, stmt, op_iter, SSA_OP_DEF)
    {
      FOR_EACH_IMM_USE_FAST (use_p, imm_iter, DEF_FROM_PTR (def_p))
        {
          basic_block bb = gimple_bb (USE_STMT (use_p));
          if (!flow_bb_inside_loop_p (loop, bb))
            {
              if (vect_print_dump_info (REPORT_DETAILS))
                fprintf (vect_dump, "vec_stmt_relevant_p: used out of loop.");

              if (is_gimple_debug (USE_STMT (use_p)))
                continue;

              /* We expect all such uses to be in the loop exit phis
                 (because of loop closed form)   */
              gcc_assert (gimple_code (USE_STMT (use_p)) == GIMPLE_PHI);
              gcc_assert (bb == single_exit (loop)->dest);

              *live_p = true;
            }
        }
    }

  return (*live_p || *relevant);
}


/* Function exist_non_indexing_operands_for_use_p

   USE is one of the uses attached to STMT.  Check if USE is
   used in STMT for anything other than indexing an array.  */

static bool
exist_non_indexing_operands_for_use_p (tree use, gimple stmt)
{
  tree operand;
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt);

  /* USE corresponds to some operand in STMT.  If there is no data
     reference in STMT, then any operand that corresponds to USE
     is not indexing an array.  */
  if (!STMT_VINFO_DATA_REF (stmt_info))
    return true;

  /* STMT has a data_ref. FORNOW this means that its of one of
     the following forms:
     -1- ARRAY_REF = var
     -2- var = ARRAY_REF
     (This should have been verified in analyze_data_refs).

     'var' in the second case corresponds to a def, not a use,
     so USE cannot correspond to any operands that are not used
     for array indexing.

     Therefore, all we need to check is if STMT falls into the
     first case, and whether var corresponds to USE.  */

  if (!gimple_assign_copy_p (stmt))
    return false;
  if (TREE_CODE (gimple_assign_lhs (stmt)) == SSA_NAME)
    return false;
  operand = gimple_assign_rhs1 (stmt);
  if (TREE_CODE (operand) != SSA_NAME)
    return false;

  if (operand == use)
    return true;

  return false;
}


/*
   Function process_use.

   Inputs:
   - a USE in STMT in a loop represented by LOOP_VINFO
   - LIVE_P, RELEVANT - enum values to be set in the STMT_VINFO of the stmt
     that defined USE.  This is done by calling mark_relevant and passing it
     the WORKLIST (to add DEF_STMT to the WORKLIST in case it is relevant).

   Outputs:
   Generally, LIVE_P and RELEVANT are used to define the liveness and
   relevance info of the DEF_STMT of this USE:
       STMT_VINFO_LIVE_P (DEF_STMT_info) <-- live_p
       STMT_VINFO_RELEVANT (DEF_STMT_info) <-- relevant
   Exceptions:
   - case 1: If USE is used only for address computations (e.g. array indexing),
   which does not need to be directly vectorized, then the liveness/relevance
   of the respective DEF_STMT is left unchanged.
   - case 2: If STMT is a reduction phi and DEF_STMT is a reduction stmt, we
   skip DEF_STMT cause it had already been processed.
   - case 3: If DEF_STMT and STMT are in different nests, then  "relevant" will
   be modified accordingly.

   Return true if everything is as expected. Return false otherwise.  */

static bool
process_use (gimple stmt, tree use, loop_vec_info loop_vinfo, bool live_p,
             enum vect_relevant relevant, VEC(gimple,heap) **worklist)
{
  struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
  stmt_vec_info stmt_vinfo = vinfo_for_stmt (stmt);
  stmt_vec_info dstmt_vinfo;
  basic_block bb, def_bb;
  tree def;
  gimple def_stmt;
  enum vect_def_type dt;

  /* case 1: we are only interested in uses that need to be vectorized.  Uses
     that are used for address computation are not considered relevant.  */
  if (!exist_non_indexing_operands_for_use_p (use, stmt))
     return true;

  if (!vect_is_simple_use (use, loop_vinfo, NULL, &def_stmt, &def, &dt))
    {
      if (vect_print_dump_info (REPORT_UNVECTORIZED_LOCATIONS))
        fprintf (vect_dump, "not vectorized: unsupported use in stmt.");
      return false;
    }

  if (!def_stmt || gimple_nop_p (def_stmt))
    return true;

  def_bb = gimple_bb (def_stmt);
  if (!flow_bb_inside_loop_p (loop, def_bb))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "def_stmt is out of loop.");
      return true;
    }

  /* case 2: A reduction phi (STMT) defined by a reduction stmt (DEF_STMT).
     DEF_STMT must have already been processed, because this should be the
     only way that STMT, which is a reduction-phi, was put in the worklist,
     as there should be no other uses for DEF_STMT in the loop.  So we just
     check that everything is as expected, and we are done.  */
  dstmt_vinfo = vinfo_for_stmt (def_stmt);
  bb = gimple_bb (stmt);
  if (gimple_code (stmt) == GIMPLE_PHI
      && STMT_VINFO_DEF_TYPE (stmt_vinfo) == vect_reduction_def
      && gimple_code (def_stmt) != GIMPLE_PHI
      && STMT_VINFO_DEF_TYPE (dstmt_vinfo) == vect_reduction_def
      && bb->loop_father == def_bb->loop_father)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "reduc-stmt defining reduc-phi in the same nest.");
      if (STMT_VINFO_IN_PATTERN_P (dstmt_vinfo))
        dstmt_vinfo = vinfo_for_stmt (STMT_VINFO_RELATED_STMT (dstmt_vinfo));
      gcc_assert (STMT_VINFO_RELEVANT (dstmt_vinfo) < vect_used_by_reduction);
      gcc_assert (STMT_VINFO_LIVE_P (dstmt_vinfo)
                  || STMT_VINFO_RELEVANT (dstmt_vinfo) > vect_unused_in_scope);
      return true;
    }

  /* case 3a: outer-loop stmt defining an inner-loop stmt:
        outer-loop-header-bb:
                d = def_stmt
        inner-loop:
                stmt # use (d)
        outer-loop-tail-bb:
                ...               */
  if (flow_loop_nested_p (def_bb->loop_father, bb->loop_father))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "outer-loop def-stmt defining inner-loop stmt.");

      switch (relevant)
        {
        case vect_unused_in_scope:
          relevant = (STMT_VINFO_DEF_TYPE (stmt_vinfo) == vect_nested_cycle) ?
                      vect_used_in_scope : vect_unused_in_scope;
          break;

        case vect_used_in_outer_by_reduction:
          gcc_assert (STMT_VINFO_DEF_TYPE (stmt_vinfo) != vect_reduction_def);
          relevant = vect_used_by_reduction;
          break;

        case vect_used_in_outer:
          gcc_assert (STMT_VINFO_DEF_TYPE (stmt_vinfo) != vect_reduction_def);
          relevant = vect_used_in_scope;
          break;

        case vect_used_in_scope:
          break;

        default:
          gcc_unreachable ();
        }
    }

  /* case 3b: inner-loop stmt defining an outer-loop stmt:
        outer-loop-header-bb:
                ...
        inner-loop:
                d = def_stmt
        outer-loop-tail-bb (or outer-loop-exit-bb in double reduction):
                stmt # use (d)          */
  else if (flow_loop_nested_p (bb->loop_father, def_bb->loop_father))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "inner-loop def-stmt defining outer-loop stmt.");

      switch (relevant)
        {
        case vect_unused_in_scope:
          relevant = (STMT_VINFO_DEF_TYPE (stmt_vinfo) == vect_reduction_def
            || STMT_VINFO_DEF_TYPE (stmt_vinfo) == vect_double_reduction_def) ?
                      vect_used_in_outer_by_reduction : vect_unused_in_scope;
          break;

        case vect_used_by_reduction:
          relevant = vect_used_in_outer_by_reduction;
          break;

        case vect_used_in_scope:
          relevant = vect_used_in_outer;
          break;

        default:
          gcc_unreachable ();
        }
    }

  vect_mark_relevant (worklist, def_stmt, relevant, live_p,
                      is_pattern_stmt_p (stmt_vinfo));
  return true;
}


/* Function vect_mark_stmts_to_be_vectorized.

   Not all stmts in the loop need to be vectorized. For example:

     for i...
       for j...
   1.    T0 = i + j
   2.    T1 = a[T0]

   3.    j = j + 1

   Stmt 1 and 3 do not need to be vectorized, because loop control and
   addressing of vectorized data-refs are handled differently.

   This pass detects such stmts.  */

bool
vect_mark_stmts_to_be_vectorized (loop_vec_info loop_vinfo)
{
  VEC(gimple,heap) *worklist;
  struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
  basic_block *bbs = LOOP_VINFO_BBS (loop_vinfo);
  unsigned int nbbs = loop->num_nodes;
  gimple_stmt_iterator si;
  gimple stmt;
  unsigned int i;
  stmt_vec_info stmt_vinfo;
  basic_block bb;
  gimple phi;
  bool live_p;
  enum vect_relevant relevant, tmp_relevant;
  enum vect_def_type def_type;

  if (vect_print_dump_info (REPORT_DETAILS))
    fprintf (vect_dump, "=== vect_mark_stmts_to_be_vectorized ===");

  worklist = VEC_alloc (gimple, heap, 64);

  /* 1. Init worklist.  */
  for (i = 0; i < nbbs; i++)
    {
      bb = bbs[i];
      for (si = gsi_start_phis (bb); !gsi_end_p (si); gsi_next (&si))
        {
          phi = gsi_stmt (si);
          if (vect_print_dump_info (REPORT_DETAILS))
            {
              fprintf (vect_dump, "init: phi relevant? ");
              print_gimple_stmt (vect_dump, phi, 0, TDF_SLIM);
            }

          if (vect_stmt_relevant_p (phi, loop_vinfo, &relevant, &live_p))
            vect_mark_relevant (&worklist, phi, relevant, live_p, false);
        }
      for (si = gsi_start_bb (bb); !gsi_end_p (si); gsi_next (&si))
        {
          stmt = gsi_stmt (si);
          if (vect_print_dump_info (REPORT_DETAILS))
            {
              fprintf (vect_dump, "init: stmt relevant? ");
              print_gimple_stmt (vect_dump, stmt, 0, TDF_SLIM);
            }

          if (vect_stmt_relevant_p (stmt, loop_vinfo, &relevant, &live_p))
            vect_mark_relevant (&worklist, stmt, relevant, live_p, false);
        }
    }

  /* 2. Process_worklist */
  while (VEC_length (gimple, worklist) > 0)
    {
      use_operand_p use_p;
      ssa_op_iter iter;

      stmt = VEC_pop (gimple, worklist);
      if (vect_print_dump_info (REPORT_DETAILS))
        {
          fprintf (vect_dump, "worklist: examine stmt: ");
          print_gimple_stmt (vect_dump, stmt, 0, TDF_SLIM);
        }

      /* Examine the USEs of STMT. For each USE, mark the stmt that defines it
         (DEF_STMT) as relevant/irrelevant and live/dead according to the
         liveness and relevance properties of STMT.  */
      stmt_vinfo = vinfo_for_stmt (stmt);
      relevant = STMT_VINFO_RELEVANT (stmt_vinfo);
      live_p = STMT_VINFO_LIVE_P (stmt_vinfo);

      /* Generally, the liveness and relevance properties of STMT are
         propagated as is to the DEF_STMTs of its USEs:
          live_p <-- STMT_VINFO_LIVE_P (STMT_VINFO)
          relevant <-- STMT_VINFO_RELEVANT (STMT_VINFO)

         One exception is when STMT has been identified as defining a reduction
         variable; in this case we set the liveness/relevance as follows:
           live_p = false
           relevant = vect_used_by_reduction
         This is because we distinguish between two kinds of relevant stmts -
         those that are used by a reduction computation, and those that are
         (also) used by a regular computation.  This allows us later on to
         identify stmts that are used solely by a reduction, and therefore the
         order of the results that they produce does not have to be kept.  */

      def_type = STMT_VINFO_DEF_TYPE (stmt_vinfo);
      tmp_relevant = relevant;
      switch (def_type)
        {
          case vect_reduction_def:
            switch (tmp_relevant)
              {
                case vect_unused_in_scope:
                  relevant = vect_used_by_reduction;
                  break;

                case vect_used_by_reduction:
                  if (gimple_code (stmt) == GIMPLE_PHI)
                    break;
                  /* fall through */

                default:
                  if (vect_print_dump_info (REPORT_DETAILS))
                    fprintf (vect_dump, "unsupported use of reduction.");

                  VEC_free (gimple, heap, worklist);
                  return false;
              }

            live_p = false;
            break;

          case vect_nested_cycle:
            if (tmp_relevant != vect_unused_in_scope
                && tmp_relevant != vect_used_in_outer_by_reduction
                && tmp_relevant != vect_used_in_outer)
              {
                if (vect_print_dump_info (REPORT_DETAILS))
                  fprintf (vect_dump, "unsupported use of nested cycle.");

                VEC_free (gimple, heap, worklist);
                return false;
              }

            live_p = false;
            break;

          case vect_double_reduction_def:
            if (tmp_relevant != vect_unused_in_scope
                && tmp_relevant != vect_used_by_reduction)
              {
                if (vect_print_dump_info (REPORT_DETAILS))
                  fprintf (vect_dump, "unsupported use of double reduction.");

                VEC_free (gimple, heap, worklist);
                return false;
              }

            live_p = false;
            break;

          default:
            break;
        }

      if (is_pattern_stmt_p (vinfo_for_stmt (stmt)))
        {
          /* Pattern statements are not inserted into the code, so
             FOR_EACH_PHI_OR_STMT_USE optimizes their operands out, and we
             have to scan the RHS or function arguments instead.  */
          if (is_gimple_assign (stmt))
            {
              enum tree_code rhs_code = gimple_assign_rhs_code (stmt);
              tree op = gimple_assign_rhs1 (stmt);

              i = 1;
              if (rhs_code == COND_EXPR && COMPARISON_CLASS_P (op))
                {
                  if (!process_use (stmt, TREE_OPERAND (op, 0), loop_vinfo,
                                    live_p, relevant, &worklist)
                      || !process_use (stmt, TREE_OPERAND (op, 1), loop_vinfo,
                                       live_p, relevant, &worklist))
                    {
                      VEC_free (gimple, heap, worklist);
                      return false;
                    }
                  i = 2;
                }
              for (; i < gimple_num_ops (stmt); i++)
                {
                  op = gimple_op (stmt, i);
                  if (!process_use (stmt, op, loop_vinfo, live_p, relevant,
                                    &worklist))
                    {
                      VEC_free (gimple, heap, worklist);
                      return false;
                    }
                 }
            }
          else if (is_gimple_call (stmt))
            {
              for (i = 0; i < gimple_call_num_args (stmt); i++)
                {
                  tree arg = gimple_call_arg (stmt, i);
                  if (!process_use (stmt, arg, loop_vinfo, live_p, relevant,
                                    &worklist))
                    {
                      VEC_free (gimple, heap, worklist);
                      return false;
                    }
                }
            }
        }
      else
        FOR_EACH_PHI_OR_STMT_USE (use_p, stmt, iter, SSA_OP_USE)
          {
            tree op = USE_FROM_PTR (use_p);
            if (!process_use (stmt, op, loop_vinfo, live_p, relevant,
                              &worklist))
              {
                VEC_free (gimple, heap, worklist);
                return false;
              }
          }
    } /* while worklist */

  VEC_free (gimple, heap, worklist);
  return true;
}


/* Get cost by calling cost target builtin.  */

static inline
int vect_get_stmt_cost (enum vect_cost_for_stmt type_of_cost)
{
  tree dummy_type = NULL;
  int dummy = 0;

  return targetm.vectorize.builtin_vectorization_cost (type_of_cost,
                                                       dummy_type, dummy);
}


/* Get cost for STMT.  */

int
cost_for_stmt (gimple stmt)
{
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt);

  switch (STMT_VINFO_TYPE (stmt_info))
  {
  case load_vec_info_type:
    return vect_get_stmt_cost (scalar_load);
  case store_vec_info_type:
    return vect_get_stmt_cost (scalar_store);
  case op_vec_info_type:
  case condition_vec_info_type:
  case assignment_vec_info_type:
  case reduc_vec_info_type:
  case induc_vec_info_type:
  case type_promotion_vec_info_type:
  case type_demotion_vec_info_type:
  case type_conversion_vec_info_type:
  case call_vec_info_type:
    return vect_get_stmt_cost (scalar_stmt);
  case undef_vec_info_type:
  default:
    gcc_unreachable ();
  }
}

/* Function vect_model_simple_cost.

   Models cost for simple operations, i.e. those that only emit ncopies of a
   single op.  Right now, this does not account for multiple insns that could
   be generated for the single vector op.  We will handle that shortly.  */

void
vect_model_simple_cost (stmt_vec_info stmt_info, int ncopies,
                        enum vect_def_type *dt, slp_tree slp_node)
{
  int i;
  int inside_cost = 0, outside_cost = 0;

  /* The SLP costs were already calculated during SLP tree build.  */
  if (PURE_SLP_STMT (stmt_info))
    return;

  inside_cost = ncopies * vect_get_stmt_cost (vector_stmt); 

  /* FORNOW: Assuming maximum 2 args per stmts.  */
  for (i = 0; i < 2; i++)
    {
      if (dt[i] == vect_constant_def || dt[i] == vect_external_def)
        outside_cost += vect_get_stmt_cost (vector_stmt); 
    }

  if (vect_print_dump_info (REPORT_COST))
    fprintf (vect_dump, "vect_model_simple_cost: inside_cost = %d, "
             "outside_cost = %d .", inside_cost, outside_cost);

  /* Set the costs either in STMT_INFO or SLP_NODE (if exists).  */
  stmt_vinfo_set_inside_of_loop_cost (stmt_info, slp_node, inside_cost);
  stmt_vinfo_set_outside_of_loop_cost (stmt_info, slp_node, outside_cost);
}


/* Function vect_cost_strided_group_size

   For strided load or store, return the group_size only if it is the first
   load or store of a group, else return 1.  This ensures that group size is
   only returned once per group.  */

static int
vect_cost_strided_group_size (stmt_vec_info stmt_info)
{
  gimple first_stmt = GROUP_FIRST_ELEMENT (stmt_info);

  if (first_stmt == STMT_VINFO_STMT (stmt_info))
    return GROUP_SIZE (stmt_info);

  return 1;
}


/* Function vect_model_store_cost

   Models cost for stores.  In the case of strided accesses, one access
   has the overhead of the strided access attributed to it.  */

void
vect_model_store_cost (stmt_vec_info stmt_info, int ncopies,
                       bool store_lanes_p, enum vect_def_type dt,
                       slp_tree slp_node)
{
  int group_size;
  unsigned int inside_cost = 0, outside_cost = 0;
  struct data_reference *first_dr;
  gimple first_stmt;

  /* The SLP costs were already calculated during SLP tree build.  */
  if (PURE_SLP_STMT (stmt_info))
    return;

  if (dt == vect_constant_def || dt == vect_external_def)
    outside_cost = vect_get_stmt_cost (scalar_to_vec); 

  /* Strided access?  */
  if (STMT_VINFO_STRIDED_ACCESS (stmt_info))
    {
      if (slp_node)
        {
          first_stmt = VEC_index (gimple, SLP_TREE_SCALAR_STMTS (slp_node), 0);
          group_size = 1;
        }
      else
        {
          first_stmt = GROUP_FIRST_ELEMENT (stmt_info);
          group_size = vect_cost_strided_group_size (stmt_info);
        }

      first_dr = STMT_VINFO_DATA_REF (vinfo_for_stmt (first_stmt));
    }
  /* Not a strided access.  */
  else
    {
      group_size = 1;
      first_dr = STMT_VINFO_DATA_REF (stmt_info);
    }

  /* We assume that the cost of a single store-lanes instruction is
     equivalent to the cost of GROUP_SIZE separate stores.  If a strided
     access is instead being provided by a permute-and-store operation,
     include the cost of the permutes.  */
  if (!store_lanes_p && group_size > 1)
    {
      /* Uses a high and low interleave operation for each needed permute.  */
      inside_cost = ncopies * exact_log2(group_size) * group_size
        * vect_get_stmt_cost (vector_stmt);

      if (vect_print_dump_info (REPORT_COST))
        fprintf (vect_dump, "vect_model_store_cost: strided group_size = %d .",
                 group_size);

    }

  /* Costs of the stores.  */
  vect_get_store_cost (first_dr, ncopies, &inside_cost);

  if (vect_print_dump_info (REPORT_COST))
    fprintf (vect_dump, "vect_model_store_cost: inside_cost = %d, "
             "outside_cost = %d .", inside_cost, outside_cost);

  /* Set the costs either in STMT_INFO or SLP_NODE (if exists).  */
  stmt_vinfo_set_inside_of_loop_cost (stmt_info, slp_node, inside_cost);
  stmt_vinfo_set_outside_of_loop_cost (stmt_info, slp_node, outside_cost);
}


/* Calculate cost of DR's memory access.  */
void
vect_get_store_cost (struct data_reference *dr, int ncopies,
                     unsigned int *inside_cost)
{
  int alignment_support_scheme = vect_supportable_dr_alignment (dr, false);

  switch (alignment_support_scheme)
    {
    case dr_aligned:
      {
        *inside_cost += ncopies * vect_get_stmt_cost (vector_store);

        if (vect_print_dump_info (REPORT_COST))
          fprintf (vect_dump, "vect_model_store_cost: aligned.");

        break;
      }

    case dr_unaligned_supported:
      {
        gimple stmt = DR_STMT (dr);
        stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
        tree vectype = STMT_VINFO_VECTYPE (stmt_info);

        /* Here, we assign an additional cost for the unaligned store.  */
        *inside_cost += ncopies
          * targetm.vectorize.builtin_vectorization_cost (unaligned_store,
                                 vectype, DR_MISALIGNMENT (dr));

        if (vect_print_dump_info (REPORT_COST))
          fprintf (vect_dump, "vect_model_store_cost: unaligned supported by "
                   "hardware.");

        break;
      }

    default:
      gcc_unreachable ();
    }
}


/* Function vect_model_load_cost

   Models cost for loads.  In the case of strided accesses, the last access
   has the overhead of the strided access attributed to it.  Since unaligned
   accesses are supported for loads, we also account for the costs of the
   access scheme chosen.  */

void
vect_model_load_cost (stmt_vec_info stmt_info, int ncopies, bool load_lanes_p,
                      slp_tree slp_node)
{
  int group_size;
  gimple first_stmt;
  struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info), *first_dr;
  unsigned int inside_cost = 0, outside_cost = 0;

  /* The SLP costs were already calculated during SLP tree build.  */
  if (PURE_SLP_STMT (stmt_info))
    return;

  /* Strided accesses?  */
  first_stmt = GROUP_FIRST_ELEMENT (stmt_info);
  if (STMT_VINFO_STRIDED_ACCESS (stmt_info) && first_stmt && !slp_node)
    {
      group_size = vect_cost_strided_group_size (stmt_info);
      first_dr = STMT_VINFO_DATA_REF (vinfo_for_stmt (first_stmt));
    }
  /* Not a strided access.  */
  else
    {
      group_size = 1;
      first_dr = dr;
    }

  /* We assume that the cost of a single load-lanes instruction is
     equivalent to the cost of GROUP_SIZE separate loads.  If a strided
     access is instead being provided by a load-and-permute operation,
     include the cost of the permutes.  */
  if (!load_lanes_p && group_size > 1)
    {
      /* Uses an even and odd extract operations for each needed permute.  */
      inside_cost = ncopies * exact_log2(group_size) * group_size
        * vect_get_stmt_cost (vector_stmt);

      if (vect_print_dump_info (REPORT_COST))
        fprintf (vect_dump, "vect_model_load_cost: strided group_size = %d .",
                 group_size);
    }

  /* The loads themselves.  */
  vect_get_load_cost (first_dr, ncopies,
         ((!STMT_VINFO_STRIDED_ACCESS (stmt_info)) || group_size > 1
          || slp_node),
         &inside_cost, &outside_cost);

  if (vect_print_dump_info (REPORT_COST))
    fprintf (vect_dump, "vect_model_load_cost: inside_cost = %d, "
             "outside_cost = %d .", inside_cost, outside_cost);

  /* Set the costs either in STMT_INFO or SLP_NODE (if exists).  */
  stmt_vinfo_set_inside_of_loop_cost (stmt_info, slp_node, inside_cost);
  stmt_vinfo_set_outside_of_loop_cost (stmt_info, slp_node, outside_cost);
}


/* Calculate cost of DR's memory access.  */
void
vect_get_load_cost (struct data_reference *dr, int ncopies,
                    bool add_realign_cost, unsigned int *inside_cost,
                    unsigned int *outside_cost)
{
  int alignment_support_scheme = vect_supportable_dr_alignment (dr, false);

  switch (alignment_support_scheme)
    {
    case dr_aligned:
      {
        *inside_cost += ncopies * vect_get_stmt_cost (vector_load); 

        if (vect_print_dump_info (REPORT_COST))
          fprintf (vect_dump, "vect_model_load_cost: aligned.");

        break;
      }
    case dr_unaligned_supported:
      {
        gimple stmt = DR_STMT (dr);
        stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
        tree vectype = STMT_VINFO_VECTYPE (stmt_info);

        /* Here, we assign an additional cost for the unaligned load.  */
        *inside_cost += ncopies
          * targetm.vectorize.builtin_vectorization_cost (unaligned_load,
                                           vectype, DR_MISALIGNMENT (dr));
        if (vect_print_dump_info (REPORT_COST))
          fprintf (vect_dump, "vect_model_load_cost: unaligned supported by "
                   "hardware.");

        break;
      }
    case dr_explicit_realign:
      {
        *inside_cost += ncopies * (2 * vect_get_stmt_cost (vector_load)
           + vect_get_stmt_cost (vector_stmt));

        /* FIXME: If the misalignment remains fixed across the iterations of
           the containing loop, the following cost should be added to the
           outside costs.  */
        if (targetm.vectorize.builtin_mask_for_load)
          *inside_cost += vect_get_stmt_cost (vector_stmt);

        break;
      }
    case dr_explicit_realign_optimized:
      {
        if (vect_print_dump_info (REPORT_COST))
          fprintf (vect_dump, "vect_model_load_cost: unaligned software "
                   "pipelined.");

        /* Unaligned software pipeline has a load of an address, an initial
           load, and possibly a mask operation to "prime" the loop.  However,
           if this is an access in a group of loads, which provide strided
           access, then the above cost should only be considered for one
           access in the group.  Inside the loop, there is a load op
           and a realignment op.  */

        if (add_realign_cost)
          {
            *outside_cost = 2 * vect_get_stmt_cost (vector_stmt);
            if (targetm.vectorize.builtin_mask_for_load)
              *outside_cost += vect_get_stmt_cost (vector_stmt);
          }

        *inside_cost += ncopies * (vect_get_stmt_cost (vector_load)
          + vect_get_stmt_cost (vector_stmt));
        break;
      }

    default:
      gcc_unreachable ();
    }
}


/* Function vect_init_vector.

   Insert a new stmt (INIT_STMT) that initializes a new vector variable with
   the vector elements of VECTOR_VAR.  Place the initialization at BSI if it
   is not NULL.  Otherwise, place the initialization at the loop preheader.
   Return the DEF of INIT_STMT.
   It will be used in the vectorization of STMT.  */

tree
vect_init_vector (gimple stmt, tree vector_var, tree vector_type,
                  gimple_stmt_iterator *gsi)
{
  stmt_vec_info stmt_vinfo = vinfo_for_stmt (stmt);
  tree new_var;
  gimple init_stmt;
  tree vec_oprnd;
  edge pe;
  tree new_temp;
  basic_block new_bb;

  new_var = vect_get_new_vect_var (vector_type, vect_simple_var, "cst_");
  add_referenced_var (new_var);
  init_stmt = gimple_build_assign  (new_var, vector_var);
  new_temp = make_ssa_name (new_var, init_stmt);
  gimple_assign_set_lhs (init_stmt, new_temp);

  if (gsi)
    vect_finish_stmt_generation (stmt, init_stmt, gsi);
  else
    {
      loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_vinfo);

      if (loop_vinfo)
        {
          struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);

          if (nested_in_vect_loop_p (loop, stmt))
            loop = loop->inner;

          pe = loop_preheader_edge (loop);
          new_bb = gsi_insert_on_edge_immediate (pe, init_stmt);
          gcc_assert (!new_bb);
        }
      else
       {
          bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_vinfo);
          basic_block bb;
          gimple_stmt_iterator gsi_bb_start;

          gcc_assert (bb_vinfo);
          bb = BB_VINFO_BB (bb_vinfo);
          gsi_bb_start = gsi_after_labels (bb);
          gsi_insert_before (&gsi_bb_start, init_stmt, GSI_SAME_STMT);
       }
    }

  if (vect_print_dump_info (REPORT_DETAILS))
    {
      fprintf (vect_dump, "created new init_stmt: ");
      print_gimple_stmt (vect_dump, init_stmt, 0, TDF_SLIM);
    }

  vec_oprnd = gimple_assign_lhs (init_stmt);
  return vec_oprnd;
}


/* Function vect_get_vec_def_for_operand.

   OP is an operand in STMT.  This function returns a (vector) def that will be
   used in the vectorized stmt for STMT.

   In the case that OP is an SSA_NAME which is defined in the loop, then
   STMT_VINFO_VEC_STMT of the defining stmt holds the relevant def.

   In case OP is an invariant or constant, a new stmt that creates a vector def
   needs to be introduced.  */

tree
vect_get_vec_def_for_operand (tree op, gimple stmt, tree *scalar_def)
{
  tree vec_oprnd;
  gimple vec_stmt;
  gimple def_stmt;
  stmt_vec_info def_stmt_info = NULL;
  stmt_vec_info stmt_vinfo = vinfo_for_stmt (stmt);
  unsigned int nunits;
  loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_vinfo);
  tree vec_inv;
  tree vec_cst;
  tree t = NULL_TREE;
  tree def;
  int i;
  enum vect_def_type dt;
  bool is_simple_use;
  tree vector_type;

  if (vect_print_dump_info (REPORT_DETAILS))
    {
      fprintf (vect_dump, "vect_get_vec_def_for_operand: ");
      print_generic_expr (vect_dump, op, TDF_SLIM);
    }

  is_simple_use = vect_is_simple_use (op, loop_vinfo, NULL, &def_stmt, &def,
                                      &dt);
  gcc_assert (is_simple_use);
  if (vect_print_dump_info (REPORT_DETAILS))
    {
      if (def)
        {
          fprintf (vect_dump, "def =  ");
          print_generic_expr (vect_dump, def, TDF_SLIM);
        }
      if (def_stmt)
        {
          fprintf (vect_dump, "  def_stmt =  ");
          print_gimple_stmt (vect_dump, def_stmt, 0, TDF_SLIM);
        }
    }

  switch (dt)
    {
    /* Case 1: operand is a constant.  */
    case vect_constant_def:
      {
        vector_type = get_vectype_for_scalar_type (TREE_TYPE (op));
        gcc_assert (vector_type);
        nunits = TYPE_VECTOR_SUBPARTS (vector_type);

        if (scalar_def)
          *scalar_def = op;

        /* Create 'vect_cst_ = {cst,cst,...,cst}'  */
        if (vect_print_dump_info (REPORT_DETAILS))
          fprintf (vect_dump, "Create vector_cst. nunits = %d", nunits);

        vec_cst = build_vector_from_val (vector_type,
                                         fold_convert (TREE_TYPE (vector_type),
                                                       op));
        return vect_init_vector (stmt, vec_cst, vector_type, NULL);
      }

    /* Case 2: operand is defined outside the loop - loop invariant.  */
    case vect_external_def:
      {
        vector_type = get_vectype_for_scalar_type (TREE_TYPE (def));
        gcc_assert (vector_type);
        nunits = TYPE_VECTOR_SUBPARTS (vector_type);

        if (scalar_def)
          *scalar_def = def;

        /* Create 'vec_inv = {inv,inv,..,inv}'  */
        if (vect_print_dump_info (REPORT_DETAILS))
          fprintf (vect_dump, "Create vector_inv.");

        for (i = nunits - 1; i >= 0; --i)
          {
            t = tree_cons (NULL_TREE, def, t);
          }

        /* FIXME: use build_constructor directly.  */
        vec_inv = build_constructor_from_list (vector_type, t);
        return vect_init_vector (stmt, vec_inv, vector_type, NULL);
      }

    /* Case 3: operand is defined inside the loop.  */
    case vect_internal_def:
      {
        if (scalar_def)
          *scalar_def = NULL/* FIXME tuples: def_stmt*/;

        /* Get the def from the vectorized stmt.  */
        def_stmt_info = vinfo_for_stmt (def_stmt);

        vec_stmt = STMT_VINFO_VEC_STMT (def_stmt_info);
        /* Get vectorized pattern statement.  */
        if (!vec_stmt
            && STMT_VINFO_IN_PATTERN_P (def_stmt_info)
            && !STMT_VINFO_RELEVANT (def_stmt_info))
          vec_stmt = STMT_VINFO_VEC_STMT (vinfo_for_stmt (
                       STMT_VINFO_RELATED_STMT (def_stmt_info)));
        gcc_assert (vec_stmt);
        if (gimple_code (vec_stmt) == GIMPLE_PHI)
          vec_oprnd = PHI_RESULT (vec_stmt);
        else if (is_gimple_call (vec_stmt))
          vec_oprnd = gimple_call_lhs (vec_stmt);
        else
          vec_oprnd = gimple_assign_lhs (vec_stmt);
        return vec_oprnd;
      }

    /* Case 4: operand is defined by a loop header phi - reduction  */
    case vect_reduction_def:
    case vect_double_reduction_def:
    case vect_nested_cycle:
      {
        struct loop *loop;

        gcc_assert (gimple_code (def_stmt) == GIMPLE_PHI);
        loop = (gimple_bb (def_stmt))->loop_father;

        /* Get the def before the loop  */
        op = PHI_ARG_DEF_FROM_EDGE (def_stmt, loop_preheader_edge (loop));
        return get_initial_def_for_reduction (stmt, op, scalar_def);
     }

    /* Case 5: operand is defined by loop-header phi - induction.  */
    case vect_induction_def:
      {
        gcc_assert (gimple_code (def_stmt) == GIMPLE_PHI);

        /* Get the def from the vectorized stmt.  */
        def_stmt_info = vinfo_for_stmt (def_stmt);
        vec_stmt = STMT_VINFO_VEC_STMT (def_stmt_info);
        if (gimple_code (vec_stmt) == GIMPLE_PHI)
          vec_oprnd = PHI_RESULT (vec_stmt);
        else
          vec_oprnd = gimple_get_lhs (vec_stmt);
        return vec_oprnd;
      }

    default:
      gcc_unreachable ();
    }
}


/* Function vect_get_vec_def_for_stmt_copy

   Return a vector-def for an operand.  This function is used when the
   vectorized stmt to be created (by the caller to this function) is a "copy"
   created in case the vectorized result cannot fit in one vector, and several
   copies of the vector-stmt are required.  In this case the vector-def is
   retrieved from the vector stmt recorded in the STMT_VINFO_RELATED_STMT field
   of the stmt that defines VEC_OPRND.
   DT is the type of the vector def VEC_OPRND.

   Context:
        In case the vectorization factor (VF) is bigger than the number
   of elements that can fit in a vectype (nunits), we have to generate
   more than one vector stmt to vectorize the scalar stmt.  This situation
   arises when there are multiple data-types operated upon in the loop; the
   smallest data-type determines the VF, and as a result, when vectorizing
   stmts operating on wider types we need to create 'VF/nunits' "copies" of the
   vector stmt (each computing a vector of 'nunits' results, and together
   computing 'VF' results in each iteration).  This function is called when
   vectorizing such a stmt (e.g. vectorizing S2 in the illustration below, in
   which VF=16 and nunits=4, so the number of copies required is 4):

   scalar stmt:         vectorized into:        STMT_VINFO_RELATED_STMT

   S1: x = load         VS1.0:  vx.0 = memref0      VS1.1
                        VS1.1:  vx.1 = memref1      VS1.2
                        VS1.2:  vx.2 = memref2      VS1.3
                        VS1.3:  vx.3 = memref3

   S2: z = x + ...      VSnew.0:  vz0 = vx.0 + ...  VSnew.1
                        VSnew.1:  vz1 = vx.1 + ...  VSnew.2
                        VSnew.2:  vz2 = vx.2 + ...  VSnew.3
                        VSnew.3:  vz3 = vx.3 + ...

   The vectorization of S1 is explained in vectorizable_load.
   The vectorization of S2:
        To create the first vector-stmt out of the 4 copies - VSnew.0 -
   the function 'vect_get_vec_def_for_operand' is called to
   get the relevant vector-def for each operand of S2.  For operand x it
   returns  the vector-def 'vx.0'.

        To create the remaining copies of the vector-stmt (VSnew.j), this
   function is called to get the relevant vector-def for each operand.  It is
   obtained from the respective VS1.j stmt, which is recorded in the
   STMT_VINFO_RELATED_STMT field of the stmt that defines VEC_OPRND.

        For example, to obtain the vector-def 'vx.1' in order to create the
   vector stmt 'VSnew.1', this function is called with VEC_OPRND='vx.0'.
   Given 'vx0' we obtain the stmt that defines it ('VS1.0'); from the
   STMT_VINFO_RELATED_STMT field of 'VS1.0' we obtain the next copy - 'VS1.1',
   and return its def ('vx.1').
   Overall, to create the above sequence this function will be called 3 times:
        vx.1 = vect_get_vec_def_for_stmt_copy (dt, vx.0);
        vx.2 = vect_get_vec_def_for_stmt_copy (dt, vx.1);
        vx.3 = vect_get_vec_def_for_stmt_copy (dt, vx.2);  */

tree
vect_get_vec_def_for_stmt_copy (enum vect_def_type dt, tree vec_oprnd)
{
  gimple vec_stmt_for_operand;
  stmt_vec_info def_stmt_info;

  /* Do nothing; can reuse same def.  */
  if (dt == vect_external_def || dt == vect_constant_def )
    return vec_oprnd;

  vec_stmt_for_operand = SSA_NAME_DEF_STMT (vec_oprnd);
  def_stmt_info = vinfo_for_stmt (vec_stmt_for_operand);
  gcc_assert (def_stmt_info);
  vec_stmt_for_operand = STMT_VINFO_RELATED_STMT (def_stmt_info);
  gcc_assert (vec_stmt_for_operand);
  vec_oprnd = gimple_get_lhs (vec_stmt_for_operand);
  if (gimple_code (vec_stmt_for_operand) == GIMPLE_PHI)
    vec_oprnd = PHI_RESULT (vec_stmt_for_operand);
  else
    vec_oprnd = gimple_get_lhs (vec_stmt_for_operand);
  return vec_oprnd;
}


/* Get vectorized definitions for the operands to create a copy of an original
   stmt.  See vect_get_vec_def_for_stmt_copy () for details.  */

static void
vect_get_vec_defs_for_stmt_copy (enum vect_def_type *dt,
                                 VEC(tree,heap) **vec_oprnds0,
                                 VEC(tree,heap) **vec_oprnds1)
{
  tree vec_oprnd = VEC_pop (tree, *vec_oprnds0);

  vec_oprnd = vect_get_vec_def_for_stmt_copy (dt[0], vec_oprnd);
  VEC_quick_push (tree, *vec_oprnds0, vec_oprnd);

  if (vec_oprnds1 && *vec_oprnds1)
    {
      vec_oprnd = VEC_pop (tree, *vec_oprnds1);
      vec_oprnd = vect_get_vec_def_for_stmt_copy (dt[1], vec_oprnd);
      VEC_quick_push (tree, *vec_oprnds1, vec_oprnd);
    }
}


/* Get vectorized definitions for OP0 and OP1, or SLP_NODE if it is not
   NULL.  */

static void
vect_get_vec_defs (tree op0, tree op1, gimple stmt,
                   VEC(tree,heap) **vec_oprnds0, VEC(tree,heap) **vec_oprnds1,
                   slp_tree slp_node)
{
  if (slp_node)
    vect_get_slp_defs (op0, op1, slp_node, vec_oprnds0, vec_oprnds1, -1);
  else
    {
      tree vec_oprnd;

      *vec_oprnds0 = VEC_alloc (tree, heap, 1);
      vec_oprnd = vect_get_vec_def_for_operand (op0, stmt, NULL);
      VEC_quick_push (tree, *vec_oprnds0, vec_oprnd);

      if (op1)
        {
          *vec_oprnds1 = VEC_alloc (tree, heap, 1);
          vec_oprnd = vect_get_vec_def_for_operand (op1, stmt, NULL);
          VEC_quick_push (tree, *vec_oprnds1, vec_oprnd);
        }
    }
}


/* Function vect_finish_stmt_generation.

   Insert a new stmt.  */

void
vect_finish_stmt_generation (gimple stmt, gimple vec_stmt,
                             gimple_stmt_iterator *gsi)
{
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
  loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
  bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);

  gcc_assert (gimple_code (stmt) != GIMPLE_LABEL);

  gsi_insert_before (gsi, vec_stmt, GSI_SAME_STMT);

  set_vinfo_for_stmt (vec_stmt, new_stmt_vec_info (vec_stmt, loop_vinfo,
                                                   bb_vinfo));

  if (vect_print_dump_info (REPORT_DETAILS))
    {
      fprintf (vect_dump, "add new stmt: ");
      print_gimple_stmt (vect_dump, vec_stmt, 0, TDF_SLIM);
    }

  gimple_set_location (vec_stmt, gimple_location (stmt));
}

/* Checks if CALL can be vectorized in type VECTYPE.  Returns
   a function declaration if the target has a vectorized version
   of the function, or NULL_TREE if the function cannot be vectorized.  */

tree
vectorizable_function (gimple call, tree vectype_out, tree vectype_in)
{
  tree fndecl = gimple_call_fndecl (call);

  /* We only handle functions that do not read or clobber memory -- i.e.
     const or novops ones.  */
  if (!(gimple_call_flags (call) & (ECF_CONST | ECF_NOVOPS)))
    return NULL_TREE;

  if (!fndecl
      || TREE_CODE (fndecl) != FUNCTION_DECL
      || !DECL_BUILT_IN (fndecl))
    return NULL_TREE;

  return targetm.vectorize.builtin_vectorized_function (fndecl, vectype_out,
                                                        vectype_in);
}

/* Function vectorizable_call.

   Check if STMT performs a function call that can be vectorized.
   If VEC_STMT is also passed, vectorize the STMT: create a vectorized
   stmt to replace it, put it in VEC_STMT, and insert it at BSI.
   Return FALSE if not a vectorizable STMT, TRUE otherwise.  */

static bool
vectorizable_call (gimple stmt, gimple_stmt_iterator *gsi, gimple *vec_stmt)
{
  tree vec_dest;
  tree scalar_dest;
  tree op, type;
  tree vec_oprnd0 = NULL_TREE, vec_oprnd1 = NULL_TREE;
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt), prev_stmt_info;
  tree vectype_out, vectype_in;
  int nunits_in;
  int nunits_out;
  loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
  tree fndecl, new_temp, def, rhs_type;
  gimple def_stmt;
  enum vect_def_type dt[3]
    = {vect_unknown_def_type, vect_unknown_def_type, vect_unknown_def_type};
  gimple new_stmt = NULL;
  int ncopies, j;
  VEC(tree, heap) *vargs = NULL;
  enum { NARROW, NONE, WIDEN } modifier;
  size_t i, nargs;
  tree lhs;

  /* FORNOW: unsupported in basic block SLP.  */
  gcc_assert (loop_vinfo);

  if (!STMT_VINFO_RELEVANT_P (stmt_info))
    return false;

  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
    return false;

  /* FORNOW: SLP not supported.  */
  if (STMT_SLP_TYPE (stmt_info))
    return false;

  /* Is STMT a vectorizable call?   */
  if (!is_gimple_call (stmt))
    return false;

  if (TREE_CODE (gimple_call_lhs (stmt)) != SSA_NAME)
    return false;

  if (stmt_can_throw_internal (stmt))
    return false;

  vectype_out = STMT_VINFO_VECTYPE (stmt_info);

  /* Process function arguments.  */
  rhs_type = NULL_TREE;
  vectype_in = NULL_TREE;
  nargs = gimple_call_num_args (stmt);

  /* Bail out if the function has more than three arguments, we do not have
     interesting builtin functions to vectorize with more than two arguments
     except for fma.  No arguments is also not good.  */
  if (nargs == 0 || nargs > 3)
    return false;

  for (i = 0; i < nargs; i++)
    {
      tree opvectype;

      op = gimple_call_arg (stmt, i);

      /* We can only handle calls with arguments of the same type.  */
      if (rhs_type
          && !types_compatible_p (rhs_type, TREE_TYPE (op)))
        {
          if (vect_print_dump_info (REPORT_DETAILS))
            fprintf (vect_dump, "argument types differ.");
          return false;
        }
      if (!rhs_type)
        rhs_type = TREE_TYPE (op);

      if (!vect_is_simple_use_1 (op, loop_vinfo, NULL,
                                 &def_stmt, &def, &dt[i], &opvectype))
        {
          if (vect_print_dump_info (REPORT_DETAILS))
            fprintf (vect_dump, "use not simple.");
          return false;
        }

      if (!vectype_in)
        vectype_in = opvectype;
      else if (opvectype
               && opvectype != vectype_in)
        {
          if (vect_print_dump_info (REPORT_DETAILS))
            fprintf (vect_dump, "argument vector types differ.");
          return false;
        }
    }
  /* If all arguments are external or constant defs use a vector type with
     the same size as the output vector type.  */
  if (!vectype_in)
    vectype_in = get_same_sized_vectype (rhs_type, vectype_out);
  if (vec_stmt)
    gcc_assert (vectype_in);
  if (!vectype_in)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        {
          fprintf (vect_dump, "no vectype for scalar type ");
          print_generic_expr (vect_dump, rhs_type, TDF_SLIM);
        }

      return false;
    }

  /* FORNOW */
  nunits_in = TYPE_VECTOR_SUBPARTS (vectype_in);
  nunits_out = TYPE_VECTOR_SUBPARTS (vectype_out);
  if (nunits_in == nunits_out / 2)
    modifier = NARROW;
  else if (nunits_out == nunits_in)
    modifier = NONE;
  else if (nunits_out == nunits_in / 2)
    modifier = WIDEN;
  else
    return false;

  /* For now, we only vectorize functions if a target specific builtin
     is available.  TODO -- in some cases, it might be profitable to
     insert the calls for pieces of the vector, in order to be able
     to vectorize other operations in the loop.  */
  fndecl = vectorizable_function (stmt, vectype_out, vectype_in);
  if (fndecl == NULL_TREE)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "function is not vectorizable.");

      return false;
    }

  gcc_assert (!gimple_vuse (stmt));

  if (modifier == NARROW)
    ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_out;
  else
    ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_in;

  /* Sanity check: make sure that at least one copy of the vectorized stmt
     needs to be generated.  */
  gcc_assert (ncopies >= 1);

  if (!vec_stmt) /* transformation not required.  */
    {
      STMT_VINFO_TYPE (stmt_info) = call_vec_info_type;
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "=== vectorizable_call ===");
      vect_model_simple_cost (stmt_info, ncopies, dt, NULL);
      return true;
    }

  /** Transform.  **/

  if (vect_print_dump_info (REPORT_DETAILS))
    fprintf (vect_dump, "transform call.");

  /* Handle def.  */
  scalar_dest = gimple_call_lhs (stmt);
  vec_dest = vect_create_destination_var (scalar_dest, vectype_out);

  prev_stmt_info = NULL;
  switch (modifier)
    {
    case NONE:
      for (j = 0; j < ncopies; ++j)
        {
          /* Build argument list for the vectorized call.  */
          if (j == 0)
            vargs = VEC_alloc (tree, heap, nargs);
          else
            VEC_truncate (tree, vargs, 0);

          for (i = 0; i < nargs; i++)
            {
              op = gimple_call_arg (stmt, i);
              if (j == 0)
                vec_oprnd0
                  = vect_get_vec_def_for_operand (op, stmt, NULL);
              else
                {
                  vec_oprnd0 = gimple_call_arg (new_stmt, i);
                  vec_oprnd0
                    = vect_get_vec_def_for_stmt_copy (dt[i], vec_oprnd0);
                }

              VEC_quick_push (tree, vargs, vec_oprnd0);
            }

          new_stmt = gimple_build_call_vec (fndecl, vargs);
          new_temp = make_ssa_name (vec_dest, new_stmt);
          gimple_call_set_lhs (new_stmt, new_temp);

          vect_finish_stmt_generation (stmt, new_stmt, gsi);
          mark_symbols_for_renaming (new_stmt);

          if (j == 0)
            STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
          else
            STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;

          prev_stmt_info = vinfo_for_stmt (new_stmt);
        }

      break;

    case NARROW:
      for (j = 0; j < ncopies; ++j)
        {
          /* Build argument list for the vectorized call.  */
          if (j == 0)
            vargs = VEC_alloc (tree, heap, nargs * 2);
          else
            VEC_truncate (tree, vargs, 0);

          for (i = 0; i < nargs; i++)
            {
              op = gimple_call_arg (stmt, i);
              if (j == 0)
                {
                  vec_oprnd0
                    = vect_get_vec_def_for_operand (op, stmt, NULL);
                  vec_oprnd1
                    = vect_get_vec_def_for_stmt_copy (dt[i], vec_oprnd0);
                }
              else
                {
                  vec_oprnd1 = gimple_call_arg (new_stmt, 2*i + 1);
                  vec_oprnd0
                    = vect_get_vec_def_for_stmt_copy (dt[i], vec_oprnd1);
                  vec_oprnd1
                    = vect_get_vec_def_for_stmt_copy (dt[i], vec_oprnd0);
                }

              VEC_quick_push (tree, vargs, vec_oprnd0);
              VEC_quick_push (tree, vargs, vec_oprnd1);
            }

          new_stmt = gimple_build_call_vec (fndecl, vargs);
          new_temp = make_ssa_name (vec_dest, new_stmt);
          gimple_call_set_lhs (new_stmt, new_temp);

          vect_finish_stmt_generation (stmt, new_stmt, gsi);
          mark_symbols_for_renaming (new_stmt);

          if (j == 0)
            STMT_VINFO_VEC_STMT (stmt_info) = new_stmt;
          else
            STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;

          prev_stmt_info = vinfo_for_stmt (new_stmt);
        }

      *vec_stmt = STMT_VINFO_VEC_STMT (stmt_info);

      break;

    case WIDEN:
      /* No current target implements this case.  */
      return false;
    }

  VEC_free (tree, heap, vargs);

  /* Update the exception handling table with the vector stmt if necessary.  */
  if (maybe_clean_or_replace_eh_stmt (stmt, *vec_stmt))
    gimple_purge_dead_eh_edges (gimple_bb (stmt));

  /* The call in STMT might prevent it from being removed in dce.
     We however cannot remove it here, due to the way the ssa name
     it defines is mapped to the new definition.  So just replace
     rhs of the statement with something harmless.  */

  type = TREE_TYPE (scalar_dest);
  if (is_pattern_stmt_p (stmt_info))
    lhs = gimple_call_lhs (STMT_VINFO_RELATED_STMT (stmt_info));
  else
    lhs = gimple_call_lhs (stmt);
  new_stmt = gimple_build_assign (lhs, build_zero_cst (type));
  set_vinfo_for_stmt (new_stmt, stmt_info);
  set_vinfo_for_stmt (stmt, NULL);
  STMT_VINFO_STMT (stmt_info) = new_stmt;
  gsi_replace (gsi, new_stmt, false);
  SSA_NAME_DEF_STMT (gimple_assign_lhs (new_stmt)) = new_stmt;

  return true;
}


/* Function vect_gen_widened_results_half

   Create a vector stmt whose code, type, number of arguments, and result
   variable are CODE, OP_TYPE, and VEC_DEST, and its arguments are
   VEC_OPRND0 and VEC_OPRND1.  The new vector stmt is to be inserted at BSI.
   In the case that CODE is a CALL_EXPR, this means that a call to DECL
   needs to be created (DECL is a function-decl of a target-builtin).
   STMT is the original scalar stmt that we are vectorizing.  */

static gimple
vect_gen_widened_results_half (enum tree_code code,
                               tree decl,
                               tree vec_oprnd0, tree vec_oprnd1, int op_type,
                               tree vec_dest, gimple_stmt_iterator *gsi,
                               gimple stmt)
{
  gimple new_stmt;
  tree new_temp;

  /* Generate half of the widened result:  */
  if (code == CALL_EXPR)
    {
      /* Target specific support  */
      if (op_type == binary_op)
        new_stmt = gimple_build_call (decl, 2, vec_oprnd0, vec_oprnd1);
      else
        new_stmt = gimple_build_call (decl, 1, vec_oprnd0);
      new_temp = make_ssa_name (vec_dest, new_stmt);
      gimple_call_set_lhs (new_stmt, new_temp);
    }
  else
    {
      /* Generic support */
      gcc_assert (op_type == TREE_CODE_LENGTH (code));
      if (op_type != binary_op)
        vec_oprnd1 = NULL;
      new_stmt = gimple_build_assign_with_ops (code, vec_dest, vec_oprnd0,
                                               vec_oprnd1);
      new_temp = make_ssa_name (vec_dest, new_stmt);
      gimple_assign_set_lhs (new_stmt, new_temp);
    }
  vect_finish_stmt_generation (stmt, new_stmt, gsi);

  return new_stmt;
}

/* Check if STMT performs a conversion operation, that can be vectorized.
   If VEC_STMT is also passed, vectorize the STMT: create a vectorized
   stmt to replace it, put it in VEC_STMT, and insert it at BSI.
   Return FALSE if not a vectorizable STMT, TRUE otherwise.  */

static bool
vectorizable_conversion (gimple stmt, gimple_stmt_iterator *gsi,
                         gimple *vec_stmt, slp_tree slp_node)
{
  tree vec_dest;
  tree scalar_dest;
  tree op0;
  tree vec_oprnd0 = NULL_TREE, vec_oprnd1 = NULL_TREE;
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
  loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
  enum tree_code code, code1 = ERROR_MARK, code2 = ERROR_MARK;
  tree decl1 = NULL_TREE, decl2 = NULL_TREE;
  tree new_temp;
  tree def;
  gimple def_stmt;
  enum vect_def_type dt[2] = {vect_unknown_def_type, vect_unknown_def_type};
  gimple new_stmt = NULL;
  stmt_vec_info prev_stmt_info;
  int nunits_in;
  int nunits_out;
  tree vectype_out, vectype_in;
  int ncopies, j;
  tree rhs_type;
  enum { NARROW, NONE, WIDEN } modifier;
  int i;
  VEC(tree,heap) *vec_oprnds0 = NULL;
  tree vop0;
  VEC(tree,heap) *dummy = NULL;
  int dummy_int;

  /* Is STMT a vectorizable conversion?   */

  /* FORNOW: unsupported in basic block SLP.  */
  gcc_assert (loop_vinfo);

  if (!STMT_VINFO_RELEVANT_P (stmt_info))
    return false;

  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
    return false;

  if (!is_gimple_assign (stmt))
    return false;

  if (TREE_CODE (gimple_assign_lhs (stmt)) != SSA_NAME)
    return false;

  code = gimple_assign_rhs_code (stmt);
  if (code != FIX_TRUNC_EXPR && code != FLOAT_EXPR)
    return false;

  /* Check types of lhs and rhs.  */
  scalar_dest = gimple_assign_lhs (stmt);
  vectype_out = STMT_VINFO_VECTYPE (stmt_info);

  op0 = gimple_assign_rhs1 (stmt);
  rhs_type = TREE_TYPE (op0);
  /* Check the operands of the operation.  */
  if (!vect_is_simple_use_1 (op0, loop_vinfo, NULL,
                             &def_stmt, &def, &dt[0], &vectype_in))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "use not simple.");
      return false;
    }
  /* If op0 is an external or constant defs use a vector type of
     the same size as the output vector type.  */
  if (!vectype_in)
    vectype_in = get_same_sized_vectype (rhs_type, vectype_out);
  if (vec_stmt)
    gcc_assert (vectype_in);
  if (!vectype_in)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        {
          fprintf (vect_dump, "no vectype for scalar type ");
          print_generic_expr (vect_dump, rhs_type, TDF_SLIM);
        }

      return false;
    }

  /* FORNOW */
  nunits_in = TYPE_VECTOR_SUBPARTS (vectype_in);
  nunits_out = TYPE_VECTOR_SUBPARTS (vectype_out);
  if (nunits_in == nunits_out / 2)
    modifier = NARROW;
  else if (nunits_out == nunits_in)
    modifier = NONE;
  else if (nunits_out == nunits_in / 2)
    modifier = WIDEN;
  else
    return false;

  if (modifier == NARROW)
    ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_out;
  else
    ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_in;

  /* Multiple types in SLP are handled by creating the appropriate number of
     vectorized stmts for each SLP node.  Hence, NCOPIES is always 1 in
     case of SLP.  */
  if (slp_node || PURE_SLP_STMT (stmt_info))
    ncopies = 1;

  /* Sanity check: make sure that at least one copy of the vectorized stmt
     needs to be generated.  */
  gcc_assert (ncopies >= 1);

  /* Supportable by target?  */
  if ((modifier == NONE
       && !supportable_convert_operation (code, vectype_out, vectype_in, &decl1, &code1))
      || (modifier == WIDEN
          && !supportable_widening_operation (code, stmt,
                                              vectype_out, vectype_in,
                                              &decl1, &decl2,
                                              &code1, &code2,
                                              &dummy_int, &dummy))
      || (modifier == NARROW
          && !supportable_narrowing_operation (code, vectype_out, vectype_in,
                                               &code1, &dummy_int, &dummy)))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "conversion not supported by target.");
      return false;
    }

  if (modifier != NONE)
    {
      /* FORNOW: SLP not supported.  */
      if (STMT_SLP_TYPE (stmt_info))
        return false;
    }

  if (!vec_stmt)                /* transformation not required.  */
    {
      STMT_VINFO_TYPE (stmt_info) = type_conversion_vec_info_type;
      return true;
    }

  /** Transform.  **/
  if (vect_print_dump_info (REPORT_DETAILS))
    fprintf (vect_dump, "transform conversion.");

  /* Handle def.  */
  vec_dest = vect_create_destination_var (scalar_dest, vectype_out);

  if (modifier == NONE && !slp_node)
    vec_oprnds0 = VEC_alloc (tree, heap, 1);

  prev_stmt_info = NULL;
  switch (modifier)
    {
    case NONE:
      for (j = 0; j < ncopies; j++)
        {
          if (j == 0)
            vect_get_vec_defs (op0, NULL, stmt, &vec_oprnds0, NULL, slp_node);
          else
            vect_get_vec_defs_for_stmt_copy (dt, &vec_oprnds0, NULL);

          FOR_EACH_VEC_ELT (tree, vec_oprnds0, i, vop0)
          {
            /* Arguments are ready, create the new vector stmt.  */
            if (code1 == CALL_EXPR)
              {
                new_stmt = gimple_build_call (decl1, 1, vop0);
                new_temp = make_ssa_name (vec_dest, new_stmt);
                gimple_call_set_lhs (new_stmt, new_temp);
              }
            else
              {
                gcc_assert (TREE_CODE_LENGTH (code) == unary_op);
                new_stmt = gimple_build_assign_with_ops (code, vec_dest, vop0,
                                                        NULL);
                new_temp = make_ssa_name (vec_dest, new_stmt);
                gimple_assign_set_lhs (new_stmt, new_temp);
              }

            vect_finish_stmt_generation (stmt, new_stmt, gsi);
            if (slp_node)
              VEC_quick_push (gimple, SLP_TREE_VEC_STMTS (slp_node), new_stmt);
          }

          if (j == 0)
            STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
          else
            STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
          prev_stmt_info = vinfo_for_stmt (new_stmt);
        }
      break;

    case WIDEN:
      /* In case the vectorization factor (VF) is bigger than the number
         of elements that we can fit in a vectype (nunits), we have to
         generate more than one vector stmt - i.e - we need to "unroll"
         the vector stmt by a factor VF/nunits.  */
      for (j = 0; j < ncopies; j++)
        {
          if (j == 0)
            vec_oprnd0 = vect_get_vec_def_for_operand (op0, stmt, NULL);
          else
            vec_oprnd0 = vect_get_vec_def_for_stmt_copy (dt[0], vec_oprnd0);

          /* Generate first half of the widened result:  */
          new_stmt
            = vect_gen_widened_results_half (code1, decl1,
                                             vec_oprnd0, vec_oprnd1,
                                             unary_op, vec_dest, gsi, stmt);
          if (j == 0)
            STMT_VINFO_VEC_STMT (stmt_info) = new_stmt;
          else
            STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
          prev_stmt_info = vinfo_for_stmt (new_stmt);

          /* Generate second half of the widened result:  */
          new_stmt
            = vect_gen_widened_results_half (code2, decl2,
                                             vec_oprnd0, vec_oprnd1,
                                             unary_op, vec_dest, gsi, stmt);
          STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
          prev_stmt_info = vinfo_for_stmt (new_stmt);
        }
      break;

    case NARROW:
      /* In case the vectorization factor (VF) is bigger than the number
         of elements that we can fit in a vectype (nunits), we have to
         generate more than one vector stmt - i.e - we need to "unroll"
         the vector stmt by a factor VF/nunits.  */
      for (j = 0; j < ncopies; j++)
        {
          /* Handle uses.  */
          if (j == 0)
            {
              vec_oprnd0 = vect_get_vec_def_for_operand (op0, stmt, NULL);
              vec_oprnd1 = vect_get_vec_def_for_stmt_copy (dt[0], vec_oprnd0);
            }
          else
            {
              vec_oprnd0 = vect_get_vec_def_for_stmt_copy (dt[0], vec_oprnd1);
              vec_oprnd1 = vect_get_vec_def_for_stmt_copy (dt[0], vec_oprnd0);
            }

          /* Arguments are ready.  Create the new vector stmt.  */
          new_stmt = gimple_build_assign_with_ops (code1, vec_dest, vec_oprnd0,
                                                   vec_oprnd1);
          new_temp = make_ssa_name (vec_dest, new_stmt);
          gimple_assign_set_lhs (new_stmt, new_temp);
          vect_finish_stmt_generation (stmt, new_stmt, gsi);

          if (j == 0)
            STMT_VINFO_VEC_STMT (stmt_info) = new_stmt;
          else
            STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;

          prev_stmt_info = vinfo_for_stmt (new_stmt);
        }

      *vec_stmt = STMT_VINFO_VEC_STMT (stmt_info);
    }

  if (vec_oprnds0)
    VEC_free (tree, heap, vec_oprnds0);

  return true;
}


/* Function vectorizable_assignment.

   Check if STMT performs an assignment (copy) that can be vectorized.
   If VEC_STMT is also passed, vectorize the STMT: create a vectorized
   stmt to replace it, put it in VEC_STMT, and insert it at BSI.
   Return FALSE if not a vectorizable STMT, TRUE otherwise.  */

static bool
vectorizable_assignment (gimple stmt, gimple_stmt_iterator *gsi,
                         gimple *vec_stmt, slp_tree slp_node)
{
  tree vec_dest;
  tree scalar_dest;
  tree op;
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
  tree vectype = STMT_VINFO_VECTYPE (stmt_info);
  loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
  tree new_temp;
  tree def;
  gimple def_stmt;
  enum vect_def_type dt[2] = {vect_unknown_def_type, vect_unknown_def_type};
  unsigned int nunits = TYPE_VECTOR_SUBPARTS (vectype);
  int ncopies;
  int i, j;
  VEC(tree,heap) *vec_oprnds = NULL;
  tree vop;
  bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);
  gimple new_stmt = NULL;
  stmt_vec_info prev_stmt_info = NULL;
  enum tree_code code;
  tree vectype_in;

  /* Multiple types in SLP are handled by creating the appropriate number of
     vectorized stmts for each SLP node. Hence, NCOPIES is always 1 in
     case of SLP.  */
  if (slp_node || PURE_SLP_STMT (stmt_info))
    ncopies = 1;
  else
    ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits;

  gcc_assert (ncopies >= 1);

  if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
    return false;

  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
    return false;

  /* Is vectorizable assignment?  */
  if (!is_gimple_assign (stmt))
    return false;

  scalar_dest = gimple_assign_lhs (stmt);
  if (TREE_CODE (scalar_dest) != SSA_NAME)
    return false;

  code = gimple_assign_rhs_code (stmt);
  if (gimple_assign_single_p (stmt)
      || code == PAREN_EXPR
      || CONVERT_EXPR_CODE_P (code))
    op = gimple_assign_rhs1 (stmt);
  else
    return false;

  if (code == VIEW_CONVERT_EXPR)
    op = TREE_OPERAND (op, 0);

  if (!vect_is_simple_use_1 (op, loop_vinfo, bb_vinfo,
                             &def_stmt, &def, &dt[0], &vectype_in))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "use not simple.");
      return false;
    }

  /* We can handle NOP_EXPR conversions that do not change the number
     of elements or the vector size.  */
  if ((CONVERT_EXPR_CODE_P (code)
       || code == VIEW_CONVERT_EXPR)
      && (!vectype_in
          || TYPE_VECTOR_SUBPARTS (vectype_in) != nunits
          || (GET_MODE_SIZE (TYPE_MODE (vectype))
              != GET_MODE_SIZE (TYPE_MODE (vectype_in)))))
    return false;

  /* We do not handle bit-precision changes.  */
  if ((CONVERT_EXPR_CODE_P (code)
       || code == VIEW_CONVERT_EXPR)
      && INTEGRAL_TYPE_P (TREE_TYPE (scalar_dest))
      && ((TYPE_PRECISION (TREE_TYPE (scalar_dest))
           != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (scalar_dest))))
          || ((TYPE_PRECISION (TREE_TYPE (op))
               != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (op))))))
      /* But a conversion that does not change the bit-pattern is ok.  */
      && !((TYPE_PRECISION (TREE_TYPE (scalar_dest))
            > TYPE_PRECISION (TREE_TYPE (op)))
           && TYPE_UNSIGNED (TREE_TYPE (op))))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "type conversion to/from bit-precision "
                 "unsupported.");
      return false;
    }

  if (!vec_stmt) /* transformation not required.  */
    {
      STMT_VINFO_TYPE (stmt_info) = assignment_vec_info_type;
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "=== vectorizable_assignment ===");
      vect_model_simple_cost (stmt_info, ncopies, dt, NULL);
      return true;
    }

  /** Transform.  **/
  if (vect_print_dump_info (REPORT_DETAILS))
    fprintf (vect_dump, "transform assignment.");

  /* Handle def.  */
  vec_dest = vect_create_destination_var (scalar_dest, vectype);

  /* Handle use.  */
  for (j = 0; j < ncopies; j++)
    {
      /* Handle uses.  */
      if (j == 0)
        vect_get_vec_defs (op, NULL, stmt, &vec_oprnds, NULL, slp_node);
      else
        vect_get_vec_defs_for_stmt_copy (dt, &vec_oprnds, NULL);

      /* Arguments are ready. create the new vector stmt.  */
      FOR_EACH_VEC_ELT (tree, vec_oprnds, i, vop)
       {
         if (CONVERT_EXPR_CODE_P (code)
             || code == VIEW_CONVERT_EXPR)
           vop = build1 (VIEW_CONVERT_EXPR, vectype, vop);
         new_stmt = gimple_build_assign (vec_dest, vop);
         new_temp = make_ssa_name (vec_dest, new_stmt);
         gimple_assign_set_lhs (new_stmt, new_temp);
         vect_finish_stmt_generation (stmt, new_stmt, gsi);
         if (slp_node)
           VEC_quick_push (gimple, SLP_TREE_VEC_STMTS (slp_node), new_stmt);
       }

      if (slp_node)
        continue;

      if (j == 0)
        STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
      else
        STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;

      prev_stmt_info = vinfo_for_stmt (new_stmt);
    }

  VEC_free (tree, heap, vec_oprnds);
  return true;
}


/* Return TRUE if CODE (a shift operation) is supported for SCALAR_TYPE
   either as shift by a scalar or by a vector.  */

bool
vect_supportable_shift (enum tree_code code, tree scalar_type)
{

  enum machine_mode vec_mode;
  optab optab;
  int icode;
  tree vectype;

  vectype = get_vectype_for_scalar_type (scalar_type);
  if (!vectype)
    return false;

  optab = optab_for_tree_code (code, vectype, optab_scalar);
  if (!optab
      || optab_handler (optab, TYPE_MODE (vectype)) == CODE_FOR_nothing)
    {
      optab = optab_for_tree_code (code, vectype, optab_vector);
      if (!optab
          || (optab_handler (optab, TYPE_MODE (vectype))
                      == CODE_FOR_nothing))
        return false;
    }

  vec_mode = TYPE_MODE (vectype);
  icode = (int) optab_handler (optab, vec_mode);
  if (icode == CODE_FOR_nothing)
    return false;

  return true;
}


/* Function vectorizable_shift.

   Check if STMT performs a shift operation that can be vectorized.
   If VEC_STMT is also passed, vectorize the STMT: create a vectorized
   stmt to replace it, put it in VEC_STMT, and insert it at BSI.
   Return FALSE if not a vectorizable STMT, TRUE otherwise.  */

static bool
vectorizable_shift (gimple stmt, gimple_stmt_iterator *gsi,
                    gimple *vec_stmt, slp_tree slp_node)
{
  tree vec_dest;
  tree scalar_dest;
  tree op0, op1 = NULL;
  tree vec_oprnd1 = NULL_TREE;
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
  tree vectype;
  loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
  enum tree_code code;
  enum machine_mode vec_mode;
  tree new_temp;
  optab optab;
  int icode;
  enum machine_mode optab_op2_mode;
  tree def;
  gimple def_stmt;
  enum vect_def_type dt[2] = {vect_unknown_def_type, vect_unknown_def_type};
  gimple new_stmt = NULL;
  stmt_vec_info prev_stmt_info;
  int nunits_in;
  int nunits_out;
  tree vectype_out;
  tree op1_vectype;
  int ncopies;
  int j, i;
  VEC (tree, heap) *vec_oprnds0 = NULL, *vec_oprnds1 = NULL;
  tree vop0, vop1;
  unsigned int k;
  bool scalar_shift_arg = true;
  bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);
  int vf;

  if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
    return false;

  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
    return false;

  /* Is STMT a vectorizable binary/unary operation?   */
  if (!is_gimple_assign (stmt))
    return false;

  if (TREE_CODE (gimple_assign_lhs (stmt)) != SSA_NAME)
    return false;

  code = gimple_assign_rhs_code (stmt);

  if (!(code == LSHIFT_EXPR || code == RSHIFT_EXPR || code == LROTATE_EXPR
      || code == RROTATE_EXPR))
    return false;

  scalar_dest = gimple_assign_lhs (stmt);
  vectype_out = STMT_VINFO_VECTYPE (stmt_info);
  if (TYPE_PRECISION (TREE_TYPE (scalar_dest))
      != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (scalar_dest))))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "bit-precision shifts not supported.");
      return false;
    }

  op0 = gimple_assign_rhs1 (stmt);
  if (!vect_is_simple_use_1 (op0, loop_vinfo, bb_vinfo,
                             &def_stmt, &def, &dt[0], &vectype))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "use not simple.");
      return false;
    }
  /* If op0 is an external or constant def use a vector type with
     the same size as the output vector type.  */
  if (!vectype)
    vectype = get_same_sized_vectype (TREE_TYPE (op0), vectype_out);
  if (vec_stmt)
    gcc_assert (vectype);
  if (!vectype)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        {
          fprintf (vect_dump, "no vectype for scalar type ");
          print_generic_expr (vect_dump, TREE_TYPE (op0), TDF_SLIM);
        }

      return false;
    }

  nunits_out = TYPE_VECTOR_SUBPARTS (vectype_out);
  nunits_in = TYPE_VECTOR_SUBPARTS (vectype);
  if (nunits_out != nunits_in)
    return false;

  op1 = gimple_assign_rhs2 (stmt);
  if (!vect_is_simple_use_1 (op1, loop_vinfo, bb_vinfo, &def_stmt, &def,
                             &dt[1], &op1_vectype))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "use not simple.");
      return false;
    }

  if (loop_vinfo)
    vf = LOOP_VINFO_VECT_FACTOR (loop_vinfo);
  else
    vf = 1;

  /* Multiple types in SLP are handled by creating the appropriate number of
     vectorized stmts for each SLP node.  Hence, NCOPIES is always 1 in
     case of SLP.  */
  if (slp_node || PURE_SLP_STMT (stmt_info))
    ncopies = 1;
  else
    ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_in;

  gcc_assert (ncopies >= 1);

  /* Determine whether the shift amount is a vector, or scalar.  If the
     shift/rotate amount is a vector, use the vector/vector shift optabs.  */

  if (dt[1] == vect_internal_def && !slp_node)
    scalar_shift_arg = false;
  else if (dt[1] == vect_constant_def
           || dt[1] == vect_external_def
           || dt[1] == vect_internal_def)
    {
      /* In SLP, need to check whether the shift count is the same,
         in loops if it is a constant or invariant, it is always
         a scalar shift.  */
      if (slp_node)
        {
          VEC (gimple, heap) *stmts = SLP_TREE_SCALAR_STMTS (slp_node);
          gimple slpstmt;

          FOR_EACH_VEC_ELT (gimple, stmts, k, slpstmt)
            if (!operand_equal_p (gimple_assign_rhs2 (slpstmt), op1, 0))
              scalar_shift_arg = false;
        }
    }
  else
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "operand mode requires invariant argument.");
      return false;
    }

  /* Vector shifted by vector.  */
  if (!scalar_shift_arg)
    {
      optab = optab_for_tree_code (code, vectype, optab_vector);
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "vector/vector shift/rotate found.");
      if (!op1_vectype)
        op1_vectype = get_same_sized_vectype (TREE_TYPE (op1), vectype_out);
      if (op1_vectype == NULL_TREE
          || TYPE_MODE (op1_vectype) != TYPE_MODE (vectype))
        {
          if (vect_print_dump_info (REPORT_DETAILS))
            fprintf (vect_dump, "unusable type for last operand in"
                                " vector/vector shift/rotate.");
          return false;
        }
    }
  /* See if the machine has a vector shifted by scalar insn and if not
     then see if it has a vector shifted by vector insn.  */
  else
    {
      optab = optab_for_tree_code (code, vectype, optab_scalar);
      if (optab
          && optab_handler (optab, TYPE_MODE (vectype)) != CODE_FOR_nothing)
        {
          if (vect_print_dump_info (REPORT_DETAILS))
            fprintf (vect_dump, "vector/scalar shift/rotate found.");
        }
      else
        {
          optab = optab_for_tree_code (code, vectype, optab_vector);
          if (optab
               && (optab_handler (optab, TYPE_MODE (vectype))
                      != CODE_FOR_nothing))
            {
              scalar_shift_arg = false;

              if (vect_print_dump_info (REPORT_DETAILS))
                fprintf (vect_dump, "vector/vector shift/rotate found.");

              /* Unlike the other binary operators, shifts/rotates have
                 the rhs being int, instead of the same type as the lhs,
                 so make sure the scalar is the right type if we are
                 dealing with vectors of long long/long/short/char.  */
              if (dt[1] == vect_constant_def)
                op1 = fold_convert (TREE_TYPE (vectype), op1);
              else if (!useless_type_conversion_p (TREE_TYPE (vectype),
                                                   TREE_TYPE (op1)))
                {
                  if (slp_node
                      && TYPE_MODE (TREE_TYPE (vectype))
                         != TYPE_MODE (TREE_TYPE (op1)))
                    {
                      if (vect_print_dump_info (REPORT_DETAILS))
                      fprintf (vect_dump, "unusable type for last operand in"
                                          " vector/vector shift/rotate.");
                        return false;
                    }
                  if (vec_stmt && !slp_node)
                    {
                      op1 = fold_convert (TREE_TYPE (vectype), op1);
                      op1 = vect_init_vector (stmt, op1,
                                              TREE_TYPE (vectype), NULL);
                    }
                }
            }
        }
    }

  /* Supportable by target?  */
  if (!optab)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "no optab.");
      return false;
    }
  vec_mode = TYPE_MODE (vectype);
  icode = (int) optab_handler (optab, vec_mode);
  if (icode == CODE_FOR_nothing)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "op not supported by target.");
      /* Check only during analysis.  */
      if (GET_MODE_SIZE (vec_mode) != UNITS_PER_WORD
          || (vf < vect_min_worthwhile_factor (code)
              && !vec_stmt))
        return false;
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "proceeding using word mode.");
    }

  /* Worthwhile without SIMD support?  Check only during analysis.  */
  if (!VECTOR_MODE_P (TYPE_MODE (vectype))
      && vf < vect_min_worthwhile_factor (code)
      && !vec_stmt)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "not worthwhile without SIMD support.");
      return false;
    }

  if (!vec_stmt) /* transformation not required.  */
    {
      STMT_VINFO_TYPE (stmt_info) = shift_vec_info_type;
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "=== vectorizable_shift ===");
      vect_model_simple_cost (stmt_info, ncopies, dt, NULL);
      return true;
    }

  /** Transform.  **/

  if (vect_print_dump_info (REPORT_DETAILS))
    fprintf (vect_dump, "transform binary/unary operation.");

  /* Handle def.  */
  vec_dest = vect_create_destination_var (scalar_dest, vectype);

  /* Allocate VECs for vector operands.  In case of SLP, vector operands are
     created in the previous stages of the recursion, so no allocation is
     needed, except for the case of shift with scalar shift argument.  In that
     case we store the scalar operand in VEC_OPRNDS1 for every vector stmt to
     be created to vectorize the SLP group, i.e., SLP_NODE->VEC_STMTS_SIZE.
     In case of loop-based vectorization we allocate VECs of size 1.  We
     allocate VEC_OPRNDS1 only in case of binary operation.  */
  if (!slp_node)
    {
      vec_oprnds0 = VEC_alloc (tree, heap, 1);
      vec_oprnds1 = VEC_alloc (tree, heap, 1);
    }
  else if (scalar_shift_arg)
    vec_oprnds1 = VEC_alloc (tree, heap, slp_node->vec_stmts_size);

  prev_stmt_info = NULL;
  for (j = 0; j < ncopies; j++)
    {
      /* Handle uses.  */
      if (j == 0)
        {
          if (scalar_shift_arg)
            {
              /* Vector shl and shr insn patterns can be defined with scalar
                 operand 2 (shift operand).  In this case, use constant or loop
                 invariant op1 directly, without extending it to vector mode
                 first.  */
              optab_op2_mode = insn_data[icode].operand[2].mode;
              if (!VECTOR_MODE_P (optab_op2_mode))
                {
                  if (vect_print_dump_info (REPORT_DETAILS))
                    fprintf (vect_dump, "operand 1 using scalar mode.");
                  vec_oprnd1 = op1;
                  VEC_quick_push (tree, vec_oprnds1, vec_oprnd1);
                  if (slp_node)
                    {
                      /* Store vec_oprnd1 for every vector stmt to be created
                         for SLP_NODE.  We check during the analysis that all
                         the shift arguments are the same.
                         TODO: Allow different constants for different vector
                         stmts generated for an SLP instance.  */
                      for (k = 0; k < slp_node->vec_stmts_size - 1; k++)
                        VEC_quick_push (tree, vec_oprnds1, vec_oprnd1);
                    }
                }
            }

          /* vec_oprnd1 is available if operand 1 should be of a scalar-type
             (a special case for certain kind of vector shifts); otherwise,
             operand 1 should be of a vector type (the usual case).  */
          if (vec_oprnd1)
            vect_get_vec_defs (op0, NULL_TREE, stmt, &vec_oprnds0, NULL,
                               slp_node);
          else
            vect_get_vec_defs (op0, op1, stmt, &vec_oprnds0, &vec_oprnds1,
                               slp_node);
        }
      else
        vect_get_vec_defs_for_stmt_copy (dt, &vec_oprnds0, &vec_oprnds1);

      /* Arguments are ready.  Create the new vector stmt.  */
      FOR_EACH_VEC_ELT (tree, vec_oprnds0, i, vop0)
        {
          vop1 = VEC_index (tree, vec_oprnds1, i);
          new_stmt = gimple_build_assign_with_ops (code, vec_dest, vop0, vop1);
          new_temp = make_ssa_name (vec_dest, new_stmt);
          gimple_assign_set_lhs (new_stmt, new_temp);
          vect_finish_stmt_generation (stmt, new_stmt, gsi);
          if (slp_node)
            VEC_quick_push (gimple, SLP_TREE_VEC_STMTS (slp_node), new_stmt);
        }

      if (slp_node)
        continue;

      if (j == 0)
        STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
      else
        STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
      prev_stmt_info = vinfo_for_stmt (new_stmt);
    }

  VEC_free (tree, heap, vec_oprnds0);
  VEC_free (tree, heap, vec_oprnds1);

  return true;
}


/* Function vectorizable_operation.

   Check if STMT performs a binary, unary or ternary operation that can
   be vectorized.
   If VEC_STMT is also passed, vectorize the STMT: create a vectorized
   stmt to replace it, put it in VEC_STMT, and insert it at BSI.
   Return FALSE if not a vectorizable STMT, TRUE otherwise.  */

static bool
vectorizable_operation (gimple stmt, gimple_stmt_iterator *gsi,
                        gimple *vec_stmt, slp_tree slp_node)
{
  tree vec_dest;
  tree scalar_dest;
  tree op0, op1 = NULL_TREE, op2 = NULL_TREE;
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
  tree vectype;
  loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
  enum tree_code code;
  enum machine_mode vec_mode;
  tree new_temp;
  int op_type;
  optab optab;
  int icode;
  tree def;
  gimple def_stmt;
  enum vect_def_type dt[3]
    = {vect_unknown_def_type, vect_unknown_def_type, vect_unknown_def_type};
  gimple new_stmt = NULL;
  stmt_vec_info prev_stmt_info;
  int nunits_in;
  int nunits_out;
  tree vectype_out;
  int ncopies;
  int j, i;
  VEC(tree,heap) *vec_oprnds0 = NULL, *vec_oprnds1 = NULL, *vec_oprnds2 = NULL;
  tree vop0, vop1, vop2;
  bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);
  int vf;

  if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
    return false;

  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
    return false;

  /* Is STMT a vectorizable binary/unary operation?   */
  if (!is_gimple_assign (stmt))
    return false;

  if (TREE_CODE (gimple_assign_lhs (stmt)) != SSA_NAME)
    return false;

  code = gimple_assign_rhs_code (stmt);

  /* For pointer addition, we should use the normal plus for
     the vector addition.  */
  if (code == POINTER_PLUS_EXPR)
    code = PLUS_EXPR;

  /* Support only unary or binary operations.  */
  op_type = TREE_CODE_LENGTH (code);
  if (op_type != unary_op && op_type != binary_op && op_type != ternary_op)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "num. args = %d (not unary/binary/ternary op).",
                 op_type);
      return false;
    }

  scalar_dest = gimple_assign_lhs (stmt);
  vectype_out = STMT_VINFO_VECTYPE (stmt_info);

  /* Most operations cannot handle bit-precision types without extra
     truncations.  */
  if ((TYPE_PRECISION (TREE_TYPE (scalar_dest))
       != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (scalar_dest))))
      /* Exception are bitwise binary operations.  */
      && code != BIT_IOR_EXPR
      && code != BIT_XOR_EXPR
      && code != BIT_AND_EXPR)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "bit-precision arithmetic not supported.");
      return false;
    }

  op0 = gimple_assign_rhs1 (stmt);
  if (!vect_is_simple_use_1 (op0, loop_vinfo, bb_vinfo,
                             &def_stmt, &def, &dt[0], &vectype))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "use not simple.");
      return false;
    }
  /* If op0 is an external or constant def use a vector type with
     the same size as the output vector type.  */
  if (!vectype)
    vectype = get_same_sized_vectype (TREE_TYPE (op0), vectype_out);
  if (vec_stmt)
    gcc_assert (vectype);
  if (!vectype)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        {
          fprintf (vect_dump, "no vectype for scalar type ");
          print_generic_expr (vect_dump, TREE_TYPE (op0), TDF_SLIM);
        }

      return false;
    }

  nunits_out = TYPE_VECTOR_SUBPARTS (vectype_out);
  nunits_in = TYPE_VECTOR_SUBPARTS (vectype);
  if (nunits_out != nunits_in)
    return false;

  if (op_type == binary_op || op_type == ternary_op)
    {
      op1 = gimple_assign_rhs2 (stmt);
      if (!vect_is_simple_use (op1, loop_vinfo, bb_vinfo, &def_stmt, &def,
                               &dt[1]))
        {
          if (vect_print_dump_info (REPORT_DETAILS))
            fprintf (vect_dump, "use not simple.");
          return false;
        }
    }
  if (op_type == ternary_op)
    {
      op2 = gimple_assign_rhs3 (stmt);
      if (!vect_is_simple_use (op2, loop_vinfo, bb_vinfo, &def_stmt, &def,
                               &dt[2]))
        {
          if (vect_print_dump_info (REPORT_DETAILS))
            fprintf (vect_dump, "use not simple.");
          return false;
        }
    }

  if (loop_vinfo)
    vf = LOOP_VINFO_VECT_FACTOR (loop_vinfo);
  else
    vf = 1;

  /* Multiple types in SLP are handled by creating the appropriate number of
     vectorized stmts for each SLP node.  Hence, NCOPIES is always 1 in
     case of SLP.  */
  if (slp_node || PURE_SLP_STMT (stmt_info))
    ncopies = 1;
  else
    ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_in;

  gcc_assert (ncopies >= 1);

  /* Shifts are handled in vectorizable_shift ().  */
  if (code == LSHIFT_EXPR || code == RSHIFT_EXPR || code == LROTATE_EXPR
      || code == RROTATE_EXPR)
   return false;

  optab = optab_for_tree_code (code, vectype, optab_default);

  /* Supportable by target?  */
  if (!optab)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "no optab.");
      return false;
    }
  vec_mode = TYPE_MODE (vectype);
  icode = (int) optab_handler (optab, vec_mode);
  if (icode == CODE_FOR_nothing)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "op not supported by target.");
      /* Check only during analysis.  */
      if (GET_MODE_SIZE (vec_mode) != UNITS_PER_WORD
          || (vf < vect_min_worthwhile_factor (code)
              && !vec_stmt))
        return false;
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "proceeding using word mode.");
    }

  /* Worthwhile without SIMD support?  Check only during analysis.  */
  if (!VECTOR_MODE_P (TYPE_MODE (vectype))
      && vf < vect_min_worthwhile_factor (code)
      && !vec_stmt)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "not worthwhile without SIMD support.");
      return false;
    }

  if (!vec_stmt) /* transformation not required.  */
    {
      STMT_VINFO_TYPE (stmt_info) = op_vec_info_type;
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "=== vectorizable_operation ===");
      vect_model_simple_cost (stmt_info, ncopies, dt, NULL);
      return true;
    }

  /** Transform.  **/

  if (vect_print_dump_info (REPORT_DETAILS))
    fprintf (vect_dump, "transform binary/unary operation.");

  /* Handle def.  */
  vec_dest = vect_create_destination_var (scalar_dest, vectype);

  /* Allocate VECs for vector operands.  In case of SLP, vector operands are
     created in the previous stages of the recursion, so no allocation is
     needed, except for the case of shift with scalar shift argument.  In that
     case we store the scalar operand in VEC_OPRNDS1 for every vector stmt to
     be created to vectorize the SLP group, i.e., SLP_NODE->VEC_STMTS_SIZE.
     In case of loop-based vectorization we allocate VECs of size 1.  We
     allocate VEC_OPRNDS1 only in case of binary operation.  */
  if (!slp_node)
    {
      vec_oprnds0 = VEC_alloc (tree, heap, 1);
      if (op_type == binary_op || op_type == ternary_op)
        vec_oprnds1 = VEC_alloc (tree, heap, 1);
      if (op_type == ternary_op)
        vec_oprnds2 = VEC_alloc (tree, heap, 1);
    }

  /* In case the vectorization factor (VF) is bigger than the number
     of elements that we can fit in a vectype (nunits), we have to generate
     more than one vector stmt - i.e - we need to "unroll" the
     vector stmt by a factor VF/nunits.  In doing so, we record a pointer
     from one copy of the vector stmt to the next, in the field
     STMT_VINFO_RELATED_STMT.  This is necessary in order to allow following
     stages to find the correct vector defs to be used when vectorizing
     stmts that use the defs of the current stmt.  The example below
     illustrates the vectorization process when VF=16 and nunits=4 (i.e.,
     we need to create 4 vectorized stmts):

     before vectorization:
                                RELATED_STMT    VEC_STMT
        S1:     x = memref      -               -
        S2:     z = x + 1       -               -

     step 1: vectorize stmt S1 (done in vectorizable_load. See more details
             there):
                                RELATED_STMT    VEC_STMT
        VS1_0:  vx0 = memref0   VS1_1           -
        VS1_1:  vx1 = memref1   VS1_2           -
        VS1_2:  vx2 = memref2   VS1_3           -
        VS1_3:  vx3 = memref3   -               -
        S1:     x = load        -               VS1_0
        S2:     z = x + 1       -               -

     step2: vectorize stmt S2 (done here):
        To vectorize stmt S2 we first need to find the relevant vector
        def for the first operand 'x'.  This is, as usual, obtained from
        the vector stmt recorded in the STMT_VINFO_VEC_STMT of the stmt
        that defines 'x' (S1).  This way we find the stmt VS1_0, and the
        relevant vector def 'vx0'.  Having found 'vx0' we can generate
        the vector stmt VS2_0, and as usual, record it in the
        STMT_VINFO_VEC_STMT of stmt S2.
        When creating the second copy (VS2_1), we obtain the relevant vector
        def from the vector stmt recorded in the STMT_VINFO_RELATED_STMT of
        stmt VS1_0.  This way we find the stmt VS1_1 and the relevant
        vector def 'vx1'.  Using 'vx1' we create stmt VS2_1 and record a
        pointer to it in the STMT_VINFO_RELATED_STMT of the vector stmt VS2_0.
        Similarly when creating stmts VS2_2 and VS2_3.  This is the resulting
        chain of stmts and pointers:
                                RELATED_STMT    VEC_STMT
        VS1_0:  vx0 = memref0   VS1_1           -
        VS1_1:  vx1 = memref1   VS1_2           -
        VS1_2:  vx2 = memref2   VS1_3           -
        VS1_3:  vx3 = memref3   -               -
        S1:     x = load        -               VS1_0
        VS2_0:  vz0 = vx0 + v1  VS2_1           -
        VS2_1:  vz1 = vx1 + v1  VS2_2           -
        VS2_2:  vz2 = vx2 + v1  VS2_3           -
        VS2_3:  vz3 = vx3 + v1  -               -
        S2:     z = x + 1       -               VS2_0  */

  prev_stmt_info = NULL;
  for (j = 0; j < ncopies; j++)
    {
      /* Handle uses.  */
      if (j == 0)
        {
          if (op_type == binary_op || op_type == ternary_op)
            vect_get_vec_defs (op0, op1, stmt, &vec_oprnds0, &vec_oprnds1,
                               slp_node);
          else
            vect_get_vec_defs (op0, NULL_TREE, stmt, &vec_oprnds0, NULL,
                               slp_node);
          if (op_type == ternary_op)
            {
              vec_oprnds2 = VEC_alloc (tree, heap, 1);
              VEC_quick_push (tree, vec_oprnds2,
                              vect_get_vec_def_for_operand (op2, stmt, NULL));
            }
        }
      else
        {
          vect_get_vec_defs_for_stmt_copy (dt, &vec_oprnds0, &vec_oprnds1);
          if (op_type == ternary_op)
            {
              tree vec_oprnd = VEC_pop (tree, vec_oprnds2);
              VEC_quick_push (tree, vec_oprnds2,
                              vect_get_vec_def_for_stmt_copy (dt[2],
                                                              vec_oprnd));
            }
        }

      /* Arguments are ready.  Create the new vector stmt.  */
      FOR_EACH_VEC_ELT (tree, vec_oprnds0, i, vop0)
        {
          vop1 = ((op_type == binary_op || op_type == ternary_op)
                  ? VEC_index (tree, vec_oprnds1, i) : NULL_TREE);
          vop2 = ((op_type == ternary_op)
                  ? VEC_index (tree, vec_oprnds2, i) : NULL_TREE);
          new_stmt = gimple_build_assign_with_ops3 (code, vec_dest,
                                                    vop0, vop1, vop2);
          new_temp = make_ssa_name (vec_dest, new_stmt);
          gimple_assign_set_lhs (new_stmt, new_temp);
          vect_finish_stmt_generation (stmt, new_stmt, gsi);
          if (slp_node)
            VEC_quick_push (gimple, SLP_TREE_VEC_STMTS (slp_node), new_stmt);
        }

      if (slp_node)
        continue;

      if (j == 0)
        STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
      else
        STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
      prev_stmt_info = vinfo_for_stmt (new_stmt);
    }

  VEC_free (tree, heap, vec_oprnds0);
  if (vec_oprnds1)
    VEC_free (tree, heap, vec_oprnds1);
  if (vec_oprnds2)
    VEC_free (tree, heap, vec_oprnds2);

  return true;
}


/* Get vectorized definitions for loop-based vectorization.  For the first
   operand we call vect_get_vec_def_for_operand() (with OPRND containing
   scalar operand), and for the rest we get a copy with
   vect_get_vec_def_for_stmt_copy() using the previous vector definition
   (stored in OPRND). See vect_get_vec_def_for_stmt_copy() for details.
   The vectors are collected into VEC_OPRNDS.  */

static void
vect_get_loop_based_defs (tree *oprnd, gimple stmt, enum vect_def_type dt,
                          VEC (tree, heap) **vec_oprnds, int multi_step_cvt)
{
  tree vec_oprnd;

  /* Get first vector operand.  */
  /* All the vector operands except the very first one (that is scalar oprnd)
     are stmt copies.  */
  if (TREE_CODE (TREE_TYPE (*oprnd)) != VECTOR_TYPE)
    vec_oprnd = vect_get_vec_def_for_operand (*oprnd, stmt, NULL);
  else
    vec_oprnd = vect_get_vec_def_for_stmt_copy (dt, *oprnd);

  VEC_quick_push (tree, *vec_oprnds, vec_oprnd);

  /* Get second vector operand.  */
  vec_oprnd = vect_get_vec_def_for_stmt_copy (dt, vec_oprnd);
  VEC_quick_push (tree, *vec_oprnds, vec_oprnd);

  *oprnd = vec_oprnd;

  /* For conversion in multiple steps, continue to get operands
     recursively.  */
  if (multi_step_cvt)
    vect_get_loop_based_defs (oprnd, stmt, dt, vec_oprnds,  multi_step_cvt - 1);
}


/* Create vectorized demotion statements for vector operands from VEC_OPRNDS.
   For multi-step conversions store the resulting vectors and call the function
   recursively.  */

static void
vect_create_vectorized_demotion_stmts (VEC (tree, heap) **vec_oprnds,
                                       int multi_step_cvt, gimple stmt,
                                       VEC (tree, heap) *vec_dsts,
                                       gimple_stmt_iterator *gsi,
                                       slp_tree slp_node, enum tree_code code,
                                       stmt_vec_info *prev_stmt_info)
{
  unsigned int i;
  tree vop0, vop1, new_tmp, vec_dest;
  gimple new_stmt;
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt);

  vec_dest = VEC_pop (tree, vec_dsts);

  for (i = 0; i < VEC_length (tree, *vec_oprnds); i += 2)
    {
      /* Create demotion operation.  */
      vop0 = VEC_index (tree, *vec_oprnds, i);
      vop1 = VEC_index (tree, *vec_oprnds, i + 1);
      new_stmt = gimple_build_assign_with_ops (code, vec_dest, vop0, vop1);
      new_tmp = make_ssa_name (vec_dest, new_stmt);
      gimple_assign_set_lhs (new_stmt, new_tmp);
      vect_finish_stmt_generation (stmt, new_stmt, gsi);

      if (multi_step_cvt)
        /* Store the resulting vector for next recursive call.  */
        VEC_replace (tree, *vec_oprnds, i/2, new_tmp);
      else
        {
          /* This is the last step of the conversion sequence. Store the
             vectors in SLP_NODE or in vector info of the scalar statement
             (or in STMT_VINFO_RELATED_STMT chain).  */
          if (slp_node)
            VEC_quick_push (gimple, SLP_TREE_VEC_STMTS (slp_node), new_stmt);
          else
            {
              if (!*prev_stmt_info)
                STMT_VINFO_VEC_STMT (stmt_info) = new_stmt;
              else
                STMT_VINFO_RELATED_STMT (*prev_stmt_info) = new_stmt;

              *prev_stmt_info = vinfo_for_stmt (new_stmt);
            }
        }
    }

  /* For multi-step demotion operations we first generate demotion operations
     from the source type to the intermediate types, and then combine the
     results (stored in VEC_OPRNDS) in demotion operation to the destination
     type.  */
  if (multi_step_cvt)
    {
      /* At each level of recursion we have have of the operands we had at the
         previous level.  */
      VEC_truncate (tree, *vec_oprnds, (i+1)/2);
      vect_create_vectorized_demotion_stmts (vec_oprnds, multi_step_cvt - 1,
                                             stmt, vec_dsts, gsi, slp_node,
                                             code, prev_stmt_info);
    }
}


/* Function vectorizable_type_demotion

   Check if STMT performs a binary or unary operation that involves
   type demotion, and if it can be vectorized.
   If VEC_STMT is also passed, vectorize the STMT: create a vectorized
   stmt to replace it, put it in VEC_STMT, and insert it at BSI.
   Return FALSE if not a vectorizable STMT, TRUE otherwise.  */

static bool
vectorizable_type_demotion (gimple stmt, gimple_stmt_iterator *gsi,
                            gimple *vec_stmt, slp_tree slp_node)
{
  tree vec_dest;
  tree scalar_dest;
  tree op0;
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
  loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
  enum tree_code code, code1 = ERROR_MARK;
  tree def;
  gimple def_stmt;
  enum vect_def_type dt[2] = {vect_unknown_def_type, vect_unknown_def_type};
  stmt_vec_info prev_stmt_info;
  int nunits_in;
  int nunits_out;
  tree vectype_out;
  int ncopies;
  int j, i;
  tree vectype_in;
  int multi_step_cvt = 0;
  VEC (tree, heap) *vec_oprnds0 = NULL;
  VEC (tree, heap) *vec_dsts = NULL, *interm_types = NULL, *tmp_vec_dsts = NULL;
  tree last_oprnd, intermediate_type;
  bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);

  if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
    return false;

  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
    return false;

  /* Is STMT a vectorizable type-demotion operation?  */
  if (!is_gimple_assign (stmt))
    return false;

  if (TREE_CODE (gimple_assign_lhs (stmt)) != SSA_NAME)
    return false;

  code = gimple_assign_rhs_code (stmt);
  if (!CONVERT_EXPR_CODE_P (code))
    return false;

  scalar_dest = gimple_assign_lhs (stmt);
  vectype_out = STMT_VINFO_VECTYPE (stmt_info);

  /* Check the operands of the operation.  */
  op0 = gimple_assign_rhs1 (stmt);
  if (! ((INTEGRAL_TYPE_P (TREE_TYPE (scalar_dest))
          && INTEGRAL_TYPE_P (TREE_TYPE (op0)))
         || (SCALAR_FLOAT_TYPE_P (TREE_TYPE (scalar_dest))
             && SCALAR_FLOAT_TYPE_P (TREE_TYPE (op0)))))
    return false;

  if (INTEGRAL_TYPE_P (TREE_TYPE (scalar_dest))
      && ((TYPE_PRECISION (TREE_TYPE (scalar_dest))
           != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (scalar_dest))))
          || ((TYPE_PRECISION (TREE_TYPE (op0))
               != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (op0)))))))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "type demotion to/from bit-precision unsupported.");
      return false;
    }

  if (!vect_is_simple_use_1 (op0, loop_vinfo, bb_vinfo,
                             &def_stmt, &def, &dt[0], &vectype_in))
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "use not simple.");
      return false;
    }
  /* If op0 is an external def use a vector type with the
     same size as the output vector type if possible.  */
  if (!vectype_in)
    vectype_in = get_same_sized_vectype (TREE_TYPE (op0), vectype_out);
  if (vec_stmt)
    gcc_assert (vectype_in);
  if (!vectype_in)
    {
      if (vect_print_dump_info (REPORT_DETAILS))
        {
          fprintf (vect_dump, "no vectype for scalar type ");
          print_generic_expr (vect_dump, TREE_TYPE (op0), TDF_SLIM);
        }

      return false;
    }

  nunits_in = TYPE_VECTOR_SUBPARTS (vectype_in);
  nunits_out = TYPE_VECTOR_SUBPARTS (vectype_out);
  if (nunits_in >= nunits_out)
    return false;

  /* Multiple types in SLP are handled by creating the appropriate number of
     vectorized stmts for each SLP node.  Hence, NCOPIES is always 1 in
     case of SLP.  */
  if (slp_node || PURE_SLP_STMT (stmt_info))
    ncopies = 1;
  else
    ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_out;
  gcc_assert (ncopies >= 1);

  /* Supportable by target?  */
  if (!supportable_narrowing_operation (code, vectype_out, vectype_in,
                                        &code1, &multi_step_cvt, &interm_types))
    return false;

  if (!vec_stmt) /* transformation not required.  */
    {
      STMT_VINFO_TYPE (stmt_info) = type_demotion_vec_info_type;
      if (vect_print_dump_info (REPORT_DETAILS))
        fprintf (vect_dump, "=== vectorizable_demotion ===");
      vect_model_simple_cost (stmt_info, ncopies, dt, NULL);
      return true;
    }

  /** Transform.  **/
  if (vect_print_dump_info (REPORT_DETAILS))
    fprintf (vect_dump, "transform type demotion operation. ncopies = %d.",
             ncopies);

  /* In case of multi-step demotion, we first generate demotion operations to
     the intermediate types, and then from that types to the final one.
     We create vector destinations for the intermediate type (TYPES) received
     from supportable_narrowing_operation, and store them in the correct order
     for future use in vect_create_vectorized_demotion_stmts().  */
  if (multi_step_cvt)
    vec_dsts = VEC_alloc (tree, heap, multi_step_cvt + 1);
  else
    vec_dsts = VEC_alloc (tree, heap, 1);

  vec_dest = vect_create_destination_var (scalar_dest, vectype_out);
  VEC_quick_push (tree, vec_dsts, vec_dest);

  if (multi_step_cvt)
    {
      for (i = VEC_length (tree, interm_types) - 1;
           VEC_iterate (tree, interm_types, i, intermediate_type); i--)
        {
          vec_dest = vect_create_destination_var (scalar_dest,
                                                  intermediate_type);
          VEC_quick_push (tree, vec_dsts, vec_dest);
        }
    }

  /* In case the vectorization factor (VF) is bigger than the number
     of elements that we can fit in a vectype (nunits), we have to generate
     more than one vector stmt - i.e - we need to "unroll" the
     vector stmt by a factor VF/nunits.   */
  last_oprnd = op0;
  prev_stmt_info = NULL;
  for (j = 0; j < ncopies; j++)
    {
      /* Handle uses.  */
      if (slp_node)
        vect_get_slp_defs (op0, NULL_TREE, slp_node, &vec_oprnds0, NULL, -1);
      else
        {
          VEC_free (tree, heap, vec_oprnds0);
          vec_oprnds0 = VEC_alloc (tree, heap,
                        (multi_step_cvt ? vect_pow2 (multi_step_cvt) * 2 : 2));
          vect_get_loop_based_defs (&last_oprnd, stmt, dt[0], &vec_oprnds0,
                                    vect_pow2 (multi_step_cvt) - 1);
        }

      /* Arguments are ready.  Create the new vector stmts.  */
      tmp_vec_dsts = VEC_copy (tree, heap, vec_dsts);
      vect_create_vectorized_demotion_stmts (&vec_oprnds0,
                                             multi_step_cvt, stmt, tmp_vec_dsts,
                                             gsi, slp_node, code1,
                                             &prev_stmt_info);
    }

  VEC_free (tree, heap, vec_oprnds0);
  VEC_free (tree, heap, vec_dsts);
  VEC_free (tree, heap, tmp_vec_dsts);
  VEC_free (tree, heap, interm_types);

  *vec_stmt = STMT_VINFO_VEC_STMT (stmt_info);
  return true;
}


/* Create vectorized promotion statements for vector operands from VEC_OPRNDS0
   and VEC_OPRNDS1 (for binary operations).  For multi-step conversions store
   the resulting vectors and call the function recursively.  */

static void
vect_create_vectorized_promotion_stmts (VEC (tree, heap) **vec_oprnds0,
                                        VEC (tree, heap) **vec_oprnds1,
                                        int multi_step_cvt, gimple stmt,
                                        VEC (tree, heap) *vec_dsts,
                                        gimple_stmt_iterator *gsi,
                                        slp_tree slp_node, enum tree_code code1,
                                        enum tree_code code2, tree decl1,
                                        tree decl2, int op_type,
                                        stmt_vec_info *prev_stmt_info)
{
  int i;
  tree vop0, vop1, new_tmp1, new_tmp2, vec_dest;
  gimple new_stmt1, new_stmt2;
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
  VEC (tree, heap) *vec_tmp;

  vec_dest = VEC_pop (tree, vec_dsts);
  vec_tmp = VEC_alloc (tree, heap, VEC_length (tree, *vec_oprnds0) * 2);

  FOR_EACH_VEC_ELT (tree, *vec_oprnds0, i, vop0)
    {
      if (op_type == binary_op)
        vop1 = VEC_index (tree, *vec_oprnds1, i);
      else
        vop1 = NULL_TREE;

      /* Generate the two halves of promotion operation.  */
      new_stmt1 = vect_gen_widened_results_half (code1, decl1, vop0, vop1,
                                                 op_type, vec_dest, gsi, stmt);
      new_stmt2 = vect_gen_widened_results_half (code2, decl2, vop0, vop1,
                                                 op_type, vec_dest, gsi, stmt);
      if (is_gimple_call (new_stmt1))
        {
          new_tmp1 = gimple_call_lhs (new_stmt1);
          new_tmp2 = gimple_call_lhs (new_stmt2);
        }
      else
        {
          new_tmp1 = gimple_assign_lhs (new_stmt1);
          new_tmp2 = gimple_assign_lhs (new_stmt2);
        }

      if (multi_step_cvt)
        {
          /* Store the results for the recursive call.  */
          VEC_quick_push (tree, vec_tmp, new_tmp1);
          VEC_quick_push (tree, vec_tmp, new_tmp2);
        }
      else
        {
          /* Last step of promotion sequience - store the results.  */
          if (slp_node)
            {
              VEC_quick_push (gimple, SLP_TREE_VEC_STMTS (slp_node), new_stmt1);
              VEC_quick_push (gimple, SLP_TREE_VEC_STMTS (slp_node), new_stmt2);
            }
          else
            {
              if (!*prev_stmt_info)
                STMT_VINFO_VEC_STMT (stmt_info) = new_stmt1;
              else
                STMT_VINFO_RELATED_STMT (*prev_stmt_info) = new_stmt1;

              *prev_stmt_info = vinfo_for_stmt (new_stmt1);
              STMT_VINFO_RELATED_STMT (*prev_stmt_info) = new_stmt2;
              *prev_stmt_info = vinfo_for_stmt (new_stmt2);
            }
        }
    }

  if (multi_step_cvt)
    {
      /* For multi-step promotion operation we first generate we call the
         function recurcively for every stage.  We start from the input type,
         create promotion operations to the intermediate types, and then
         create promotions to the output type.  */
      *vec_oprnds0 = VEC_copy (tree, heap, vec_tmp);
      vect_create_vectorized_promotion_stmts (vec_oprnds0, vec_oprnds1,
                                              multi_step_cvt - 1, stmt,
                                              vec_dsts, gsi, slp_node, code1,
                                              code2, decl2, decl2, op_type,
                                              prev_stmt_info);
    }

  VEC_free (tree, heap, vec_tmp);
}


/* Function vectorizable_type_promotion

   Check if STMT performs a binary or unary operation that involves
   type promotion, and if it can be vectorized.
   If VEC_STMT is also passed, vectorize the STMT: create a vectorized
   stmt to replace it, put it in VEC_STMT, and insert it at BSI.
   Return FALSE if not a vectorizable STMT, TRUE otherwise.  */

static bool
vectorizable_type_promotion (gimple stmt, gimple_stmt_iterator *gsi,
                             gimple *vec_stmt, slp_tree slp_node)
{
  tree vec_dest;
  tree scalar_dest;
  tree op0, op1 = NULL;
  tree vec_oprnd0=NULL, vec_oprnd1=NULL;
  stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
  loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
  enum tree_code code, code1 = ERROR_MARK, code2 = ERROR_MARK;
  tree decl1 = NULL_TREE, decl2 = NULL_TREE;
  int op_type;
  tree def;
  gimple def_stmt;
  enum vect_def_type dt[2] = {vect_unknown_def_type, vect_unknown_def_type};
  stmt_vec_info prev_stmt_info;
  int nunits_in;
  int nunits_out;
  tree vectype_out;
  int ncopies;
  int j, i;
  tree vectype_in;
  tree intermediate_type = NULL_TREE;