diff gcc/ada/exp_fixd.adb @ 111:04ced10e8804

gcc 7
author kono
date Fri, 27 Oct 2017 22:46:09 +0900
parents
children 84e7813d76e9
line wrap: on
line diff
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/gcc/ada/exp_fixd.adb	Fri Oct 27 22:46:09 2017 +0900
@@ -0,0 +1,2464 @@
+------------------------------------------------------------------------------
+--                                                                          --
+--                         GNAT COMPILER COMPONENTS                         --
+--                                                                          --
+--                             E X P _ F I X D                              --
+--                                                                          --
+--                                 B o d y                                  --
+--                                                                          --
+--          Copyright (C) 1992-2017, Free Software Foundation, Inc.         --
+--                                                                          --
+-- GNAT is free software;  you can  redistribute it  and/or modify it under --
+-- terms of the  GNU General Public License as published  by the Free Soft- --
+-- ware  Foundation;  either version 3,  or (at your option) any later ver- --
+-- sion.  GNAT is distributed in the hope that it will be useful, but WITH- --
+-- OUT 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  distributed with GNAT; see file COPYING3.  If not, go to --
+-- http://www.gnu.org/licenses for a complete copy of the license.          --
+--                                                                          --
+-- GNAT was originally developed  by the GNAT team at  New York University. --
+-- Extensive contributions were provided by Ada Core Technologies Inc.      --
+--                                                                          --
+------------------------------------------------------------------------------
+
+with Atree;    use Atree;
+with Checks;   use Checks;
+with Einfo;    use Einfo;
+with Exp_Util; use Exp_Util;
+with Nlists;   use Nlists;
+with Nmake;    use Nmake;
+with Restrict; use Restrict;
+with Rident;   use Rident;
+with Rtsfind;  use Rtsfind;
+with Sem;      use Sem;
+with Sem_Eval; use Sem_Eval;
+with Sem_Res;  use Sem_Res;
+with Sem_Util; use Sem_Util;
+with Sinfo;    use Sinfo;
+with Snames;   use Snames;
+with Stand;    use Stand;
+with Tbuild;   use Tbuild;
+with Uintp;    use Uintp;
+with Urealp;   use Urealp;
+
+package body Exp_Fixd is
+
+   -----------------------
+   -- Local Subprograms --
+   -----------------------
+
+   --  General note; in this unit, a number of routines are driven by the
+   --  types (Etype) of their operands. Since we are dealing with unanalyzed
+   --  expressions as they are constructed, the Etypes would not normally be
+   --  set, but the construction routines that we use in this unit do in fact
+   --  set the Etype values correctly. In addition, setting the Etype ensures
+   --  that the analyzer does not try to redetermine the type when the node
+   --  is analyzed (which would be wrong, since in the case where we set the
+   --  Treat_Fixed_As_Integer or Conversion_OK flags, it would think it was
+   --  still dealing with a normal fixed-point operation and mess it up).
+
+   function Build_Conversion
+     (N     : Node_Id;
+      Typ   : Entity_Id;
+      Expr  : Node_Id;
+      Rchk  : Boolean := False;
+      Trunc : Boolean := False) return Node_Id;
+   --  Build an expression that converts the expression Expr to type Typ,
+   --  taking the source location from Sloc (N). If the conversions involve
+   --  fixed-point types, then the Conversion_OK flag will be set so that the
+   --  resulting conversions do not get re-expanded. On return the resulting
+   --  node has its Etype set. If Rchk is set, then Do_Range_Check is set
+   --  in the resulting conversion node. If Trunc is set, then the
+   --  Float_Truncate flag is set on the conversion, which must be from
+   --  a floating-point type to an integer type.
+
+   function Build_Divide (N : Node_Id; L, R : Node_Id) return Node_Id;
+   --  Builds an N_Op_Divide node from the given left and right operand
+   --  expressions, using the source location from Sloc (N). The operands are
+   --  either both Universal_Real, in which case Build_Divide differs from
+   --  Make_Op_Divide only in that the Etype of the resulting node is set (to
+   --  Universal_Real), or they can be integer types. In this case the integer
+   --  types need not be the same, and Build_Divide converts the operand with
+   --  the smaller sized type to match the type of the other operand and sets
+   --  this as the result type. The Rounded_Result flag of the result in this
+   --  case is set from the Rounded_Result flag of node N. On return, the
+   --  resulting node is analyzed, and has its Etype set.
+
+   function Build_Double_Divide
+     (N       : Node_Id;
+      X, Y, Z : Node_Id) return Node_Id;
+   --  Returns a node corresponding to the value X/(Y*Z) using the source
+   --  location from Sloc (N). The division is rounded if the Rounded_Result
+   --  flag of N is set. The integer types of X, Y, Z may be different. On
+   --  return the resulting node is analyzed, and has its Etype set.
+
+   procedure Build_Double_Divide_Code
+     (N        : Node_Id;
+      X, Y, Z  : Node_Id;
+      Qnn, Rnn : out Entity_Id;
+      Code     : out List_Id);
+   --  Generates a sequence of code for determining the quotient and remainder
+   --  of the division X/(Y*Z), using the source location from Sloc (N).
+   --  Entities of appropriate types are allocated for the quotient and
+   --  remainder and returned in Qnn and Rnn. The result is rounded if the
+   --  Rounded_Result flag of N is set. The Etype fields of Qnn and Rnn are
+   --  appropriately set on return.
+
+   function Build_Multiply (N : Node_Id; L, R : Node_Id) return Node_Id;
+   --  Builds an N_Op_Multiply node from the given left and right operand
+   --  expressions, using the source location from Sloc (N). The operands are
+   --  either both Universal_Real, in which case Build_Multiply differs from
+   --  Make_Op_Multiply only in that the Etype of the resulting node is set (to
+   --  Universal_Real), or they can be integer types. In this case the integer
+   --  types need not be the same, and Build_Multiply chooses a type long
+   --  enough to hold the product (i.e. twice the size of the longer of the two
+   --  operand types), and both operands are converted to this type. The Etype
+   --  of the result is also set to this value. However, the result can never
+   --  overflow Integer_64, so this is the largest type that is ever generated.
+   --  On return, the resulting node is analyzed and has its Etype set.
+
+   function Build_Rem (N : Node_Id; L, R : Node_Id) return Node_Id;
+   --  Builds an N_Op_Rem node from the given left and right operand
+   --  expressions, using the source location from Sloc (N). The operands are
+   --  both integer types, which need not be the same. Build_Rem converts the
+   --  operand with the smaller sized type to match the type of the other
+   --  operand and sets this as the result type. The result is never rounded
+   --  (rem operations cannot be rounded in any case). On return, the resulting
+   --  node is analyzed and has its Etype set.
+
+   function Build_Scaled_Divide
+     (N       : Node_Id;
+      X, Y, Z : Node_Id) return Node_Id;
+   --  Returns a node corresponding to the value X*Y/Z using the source
+   --  location from Sloc (N). The division is rounded if the Rounded_Result
+   --  flag of N is set. The integer types of X, Y, Z may be different. On
+   --  return the resulting node is analyzed and has is Etype set.
+
+   procedure Build_Scaled_Divide_Code
+     (N        : Node_Id;
+      X, Y, Z  : Node_Id;
+      Qnn, Rnn : out Entity_Id;
+      Code     : out List_Id);
+   --  Generates a sequence of code for determining the quotient and remainder
+   --  of the division X*Y/Z, using the source location from Sloc (N). Entities
+   --  of appropriate types are allocated for the quotient and remainder and
+   --  returned in Qnn and Rrr. The integer types for X, Y, Z may be different.
+   --  The division is rounded if the Rounded_Result flag of N is set. The
+   --  Etype fields of Qnn and Rnn are appropriately set on return.
+
+   procedure Do_Divide_Fixed_Fixed (N : Node_Id);
+   --  Handles expansion of divide for case of two fixed-point operands
+   --  (neither of them universal), with an integer or fixed-point result.
+   --  N is the N_Op_Divide node to be expanded.
+
+   procedure Do_Divide_Fixed_Universal (N : Node_Id);
+   --  Handles expansion of divide for case of a fixed-point operand divided
+   --  by a universal real operand, with an integer or fixed-point result. N
+   --  is the N_Op_Divide node to be expanded.
+
+   procedure Do_Divide_Universal_Fixed (N : Node_Id);
+   --  Handles expansion of divide for case of a universal real operand
+   --  divided by a fixed-point operand, with an integer or fixed-point
+   --  result. N is the N_Op_Divide node to be expanded.
+
+   procedure Do_Multiply_Fixed_Fixed (N : Node_Id);
+   --  Handles expansion of multiply for case of two fixed-point operands
+   --  (neither of them universal), with an integer or fixed-point result.
+   --  N is the N_Op_Multiply node to be expanded.
+
+   procedure Do_Multiply_Fixed_Universal (N : Node_Id; Left, Right : Node_Id);
+   --  Handles expansion of multiply for case of a fixed-point operand
+   --  multiplied by a universal real operand, with an integer or fixed-
+   --  point result. N is the N_Op_Multiply node to be expanded, and
+   --  Left, Right are the operands (which may have been switched).
+
+   procedure Expand_Convert_Fixed_Static (N : Node_Id);
+   --  This routine is called where the node N is a conversion of a literal
+   --  or other static expression of a fixed-point type to some other type.
+   --  In such cases, we simply rewrite the operand as a real literal and
+   --  reanalyze. This avoids problems which would otherwise result from
+   --  attempting to build and fold expressions involving constants.
+
+   function Fpt_Value (N : Node_Id) return Node_Id;
+   --  Given an operand of fixed-point operation, return an expression that
+   --  represents the corresponding Universal_Real value. The expression
+   --  can be of integer type, floating-point type, or fixed-point type.
+   --  The expression returned is neither analyzed and resolved. The Etype
+   --  of the result is properly set (to Universal_Real).
+
+   function Integer_Literal
+     (N        : Node_Id;
+      V        : Uint;
+      Negative : Boolean := False) return Node_Id;
+   --  Given a non-negative universal integer value, build a typed integer
+   --  literal node, using the smallest applicable standard integer type. If
+   --  and only if Negative is true a negative literal is built. If V exceeds
+   --  2**63-1, the largest value allowed for perfect result set scaling
+   --  factors (see RM G.2.3(22)), then Empty is returned. The node N provides
+   --  the Sloc value for the constructed literal. The Etype of the resulting
+   --  literal is correctly set, and it is marked as analyzed.
+
+   function Real_Literal (N : Node_Id; V : Ureal) return Node_Id;
+   --  Build a real literal node from the given value, the Etype of the
+   --  returned node is set to Universal_Real, since all floating-point
+   --  arithmetic operations that we construct use Universal_Real
+
+   function Rounded_Result_Set (N : Node_Id) return Boolean;
+   --  Returns True if N is a node that contains the Rounded_Result flag
+   --  and if the flag is true or the target type is an integer type.
+
+   procedure Set_Result
+     (N     : Node_Id;
+      Expr  : Node_Id;
+      Rchk  : Boolean := False;
+      Trunc : Boolean := False);
+   --  N is the node for the current conversion, division or multiplication
+   --  operation, and Expr is an expression representing the result. Expr may
+   --  be of floating-point or integer type. If the operation result is fixed-
+   --  point, then the value of Expr is in units of small of the result type
+   --  (i.e. small's have already been dealt with). The result of the call is
+   --  to replace N by an appropriate conversion to the result type, dealing
+   --  with rounding for the decimal types case. The node is then analyzed and
+   --  resolved using the result type. If Rchk or Trunc are True, then
+   --  respectively Do_Range_Check and Float_Truncate are set in the
+   --  resulting conversion.
+
+   ----------------------
+   -- Build_Conversion --
+   ----------------------
+
+   function Build_Conversion
+     (N     : Node_Id;
+      Typ   : Entity_Id;
+      Expr  : Node_Id;
+      Rchk  : Boolean := False;
+      Trunc : Boolean := False) return Node_Id
+   is
+      Loc    : constant Source_Ptr := Sloc (N);
+      Result : Node_Id;
+      Rcheck : Boolean := Rchk;
+
+   begin
+      --  A special case, if the expression is an integer literal and the
+      --  target type is an integer type, then just retype the integer
+      --  literal to the desired target type. Don't do this if we need
+      --  a range check.
+
+      if Nkind (Expr) = N_Integer_Literal
+        and then Is_Integer_Type (Typ)
+        and then not Rchk
+      then
+         Result := Expr;
+
+      --  Cases where we end up with a conversion. Note that we do not use the
+      --  Convert_To abstraction here, since we may be decorating the resulting
+      --  conversion with Rounded_Result and/or Conversion_OK, so we want the
+      --  conversion node present, even if it appears to be redundant.
+
+      else
+         --  Remove inner conversion if both inner and outer conversions are
+         --  to integer types, since the inner one serves no purpose (except
+         --  perhaps to set rounding, so we preserve the Rounded_Result flag)
+         --  and also we preserve the range check flag on the inner operand
+
+         if Is_Integer_Type (Typ)
+           and then Is_Integer_Type (Etype (Expr))
+           and then Nkind (Expr) = N_Type_Conversion
+         then
+            Result :=
+              Make_Type_Conversion (Loc,
+                Subtype_Mark => New_Occurrence_Of (Typ, Loc),
+                Expression   => Expression (Expr));
+            Set_Rounded_Result (Result, Rounded_Result_Set (Expr));
+            Rcheck := Rcheck or Do_Range_Check (Expr);
+
+         --  For all other cases, a simple type conversion will work
+
+         else
+            Result :=
+              Make_Type_Conversion (Loc,
+                Subtype_Mark => New_Occurrence_Of (Typ, Loc),
+                Expression   => Expr);
+
+            Set_Float_Truncate (Result, Trunc);
+         end if;
+
+         --  Set Conversion_OK if either result or expression type is a
+         --  fixed-point type, since from a semantic point of view, we are
+         --  treating fixed-point values as integers at this stage.
+
+         if Is_Fixed_Point_Type (Typ)
+           or else Is_Fixed_Point_Type (Etype (Expression (Result)))
+         then
+            Set_Conversion_OK (Result);
+         end if;
+
+         --  Set Do_Range_Check if either it was requested by the caller,
+         --  or if an eliminated inner conversion had a range check.
+
+         if Rcheck then
+            Enable_Range_Check (Result);
+         else
+            Set_Do_Range_Check (Result, False);
+         end if;
+      end if;
+
+      Set_Etype (Result, Typ);
+      return Result;
+   end Build_Conversion;
+
+   ------------------
+   -- Build_Divide --
+   ------------------
+
+   function Build_Divide (N : Node_Id; L, R : Node_Id) return Node_Id is
+      Loc         : constant Source_Ptr := Sloc (N);
+      Left_Type   : constant Entity_Id  := Base_Type (Etype (L));
+      Right_Type  : constant Entity_Id  := Base_Type (Etype (R));
+      Result_Type : Entity_Id;
+      Rnode       : Node_Id;
+
+   begin
+      --  Deal with floating-point case first
+
+      if Is_Floating_Point_Type (Left_Type) then
+         pragma Assert (Left_Type = Universal_Real);
+         pragma Assert (Right_Type = Universal_Real);
+
+         Rnode := Make_Op_Divide (Loc, L, R);
+         Result_Type := Universal_Real;
+
+      --  Integer and fixed-point cases
+
+      else
+         --  An optimization. If the right operand is the literal 1, then we
+         --  can just return the left hand operand. Putting the optimization
+         --  here allows us to omit the check at the call site.
+
+         if Nkind (R) = N_Integer_Literal and then Intval (R) = 1 then
+            return L;
+         end if;
+
+         --  If left and right types are the same, no conversion needed
+
+         if Left_Type = Right_Type then
+            Result_Type := Left_Type;
+            Rnode :=
+              Make_Op_Divide (Loc,
+                Left_Opnd  => L,
+                Right_Opnd => R);
+
+         --  Use left type if it is the larger of the two
+
+         elsif Esize (Left_Type) >= Esize (Right_Type) then
+            Result_Type := Left_Type;
+            Rnode :=
+              Make_Op_Divide (Loc,
+                Left_Opnd  => L,
+                Right_Opnd => Build_Conversion (N, Left_Type, R));
+
+         --  Otherwise right type is larger of the two, us it
+
+         else
+            Result_Type := Right_Type;
+            Rnode :=
+              Make_Op_Divide (Loc,
+                Left_Opnd => Build_Conversion (N, Right_Type, L),
+                Right_Opnd => R);
+         end if;
+      end if;
+
+      --  We now have a divide node built with Result_Type set. First
+      --  set Etype of result, as required for all Build_xxx routines
+
+      Set_Etype (Rnode, Base_Type (Result_Type));
+
+      --  Set Treat_Fixed_As_Integer if operation on fixed-point type
+      --  since this is a literal arithmetic operation, to be performed
+      --  by Gigi without any consideration of small values.
+
+      if Is_Fixed_Point_Type (Result_Type) then
+         Set_Treat_Fixed_As_Integer (Rnode);
+      end if;
+
+      --  The result is rounded if the target of the operation is decimal
+      --  and Rounded_Result is set, or if the target of the operation
+      --  is an integer type.
+
+      if Is_Integer_Type (Etype (N))
+        or else Rounded_Result_Set (N)
+      then
+         Set_Rounded_Result (Rnode);
+      end if;
+
+      return Rnode;
+   end Build_Divide;
+
+   -------------------------
+   -- Build_Double_Divide --
+   -------------------------
+
+   function Build_Double_Divide
+     (N       : Node_Id;
+      X, Y, Z : Node_Id) return Node_Id
+   is
+      Y_Size : constant Nat := UI_To_Int (Esize (Etype (Y)));
+      Z_Size : constant Nat := UI_To_Int (Esize (Etype (Z)));
+      Expr   : Node_Id;
+
+   begin
+      --  If denominator fits in 64 bits, we can build the operations directly
+      --  without causing any intermediate overflow, so that's what we do.
+
+      if Nat'Max (Y_Size, Z_Size) <= 32 then
+         return
+           Build_Divide (N, X, Build_Multiply (N, Y, Z));
+
+      --  Otherwise we use the runtime routine
+
+      --    [Qnn : Interfaces.Integer_64,
+      --     Rnn : Interfaces.Integer_64;
+      --     Double_Divide (X, Y, Z, Qnn, Rnn, Round);
+      --     Qnn]
+
+      else
+         declare
+            Loc  : constant Source_Ptr := Sloc (N);
+            Qnn  : Entity_Id;
+            Rnn  : Entity_Id;
+            Code : List_Id;
+
+            pragma Warnings (Off, Rnn);
+
+         begin
+            Build_Double_Divide_Code (N, X, Y, Z, Qnn, Rnn, Code);
+            Insert_Actions (N, Code);
+            Expr := New_Occurrence_Of (Qnn, Loc);
+
+            --  Set type of result in case used elsewhere (see note at start)
+
+            Set_Etype (Expr, Etype (Qnn));
+
+            --  Set result as analyzed (see note at start on build routines)
+
+            return Expr;
+         end;
+      end if;
+   end Build_Double_Divide;
+
+   ------------------------------
+   -- Build_Double_Divide_Code --
+   ------------------------------
+
+   --  If the denominator can be computed in 64-bits, we build
+
+   --    [Nnn : constant typ := typ (X);
+   --     Dnn : constant typ := typ (Y) * typ (Z)
+   --     Qnn : constant typ := Nnn / Dnn;
+   --     Rnn : constant typ := Nnn / Dnn;
+
+   --  If the numerator cannot be computed in 64 bits, we build
+
+   --    [Qnn : typ;
+   --     Rnn : typ;
+   --     Double_Divide (X, Y, Z, Qnn, Rnn, Round);]
+
+   procedure Build_Double_Divide_Code
+     (N        : Node_Id;
+      X, Y, Z  : Node_Id;
+      Qnn, Rnn : out Entity_Id;
+      Code     : out List_Id)
+   is
+      Loc    : constant Source_Ptr := Sloc (N);
+
+      X_Size : constant Nat := UI_To_Int (Esize (Etype (X)));
+      Y_Size : constant Nat := UI_To_Int (Esize (Etype (Y)));
+      Z_Size : constant Nat := UI_To_Int (Esize (Etype (Z)));
+
+      QR_Siz : Nat;
+      QR_Typ : Entity_Id;
+
+      Nnn : Entity_Id;
+      Dnn : Entity_Id;
+
+      Quo : Node_Id;
+      Rnd : Entity_Id;
+
+   begin
+      --  Find type that will allow computation of numerator
+
+      QR_Siz := Nat'Max (X_Size, 2 * Nat'Max (Y_Size, Z_Size));
+
+      if QR_Siz <= 16 then
+         QR_Typ := Standard_Integer_16;
+      elsif QR_Siz <= 32 then
+         QR_Typ := Standard_Integer_32;
+      elsif QR_Siz <= 64 then
+         QR_Typ := Standard_Integer_64;
+
+      --  For more than 64, bits, we use the 64-bit integer defined in
+      --  Interfaces, so that it can be handled by the runtime routine.
+
+      else
+         QR_Typ := RTE (RE_Integer_64);
+      end if;
+
+      --  Define quotient and remainder, and set their Etypes, so
+      --  that they can be picked up by Build_xxx routines.
+
+      Qnn := Make_Temporary (Loc, 'S');
+      Rnn := Make_Temporary (Loc, 'R');
+
+      Set_Etype (Qnn, QR_Typ);
+      Set_Etype (Rnn, QR_Typ);
+
+      --  Case that we can compute the denominator in 64 bits
+
+      if QR_Siz <= 64 then
+
+         --  Create temporaries for numerator and denominator and set Etypes,
+         --  so that New_Occurrence_Of picks them up for Build_xxx calls.
+
+         Nnn := Make_Temporary (Loc, 'N');
+         Dnn := Make_Temporary (Loc, 'D');
+
+         Set_Etype (Nnn, QR_Typ);
+         Set_Etype (Dnn, QR_Typ);
+
+         Code := New_List (
+           Make_Object_Declaration (Loc,
+             Defining_Identifier => Nnn,
+             Object_Definition   => New_Occurrence_Of (QR_Typ, Loc),
+             Constant_Present    => True,
+             Expression => Build_Conversion (N, QR_Typ, X)),
+
+           Make_Object_Declaration (Loc,
+             Defining_Identifier => Dnn,
+             Object_Definition   => New_Occurrence_Of (QR_Typ, Loc),
+             Constant_Present    => True,
+             Expression =>
+               Build_Multiply (N,
+                 Build_Conversion (N, QR_Typ, Y),
+                 Build_Conversion (N, QR_Typ, Z))));
+
+         Quo :=
+           Build_Divide (N,
+             New_Occurrence_Of (Nnn, Loc),
+             New_Occurrence_Of (Dnn, Loc));
+
+         Set_Rounded_Result (Quo, Rounded_Result_Set (N));
+
+         Append_To (Code,
+           Make_Object_Declaration (Loc,
+             Defining_Identifier => Qnn,
+             Object_Definition   => New_Occurrence_Of (QR_Typ, Loc),
+             Constant_Present    => True,
+             Expression          => Quo));
+
+         Append_To (Code,
+           Make_Object_Declaration (Loc,
+             Defining_Identifier => Rnn,
+             Object_Definition   => New_Occurrence_Of (QR_Typ, Loc),
+             Constant_Present    => True,
+             Expression =>
+               Build_Rem (N,
+                 New_Occurrence_Of (Nnn, Loc),
+                 New_Occurrence_Of (Dnn, Loc))));
+
+      --  Case where denominator does not fit in 64 bits, so we have to
+      --  call the runtime routine to compute the quotient and remainder
+
+      else
+         Rnd := Boolean_Literals (Rounded_Result_Set (N));
+
+         Code := New_List (
+           Make_Object_Declaration (Loc,
+             Defining_Identifier => Qnn,
+             Object_Definition   => New_Occurrence_Of (QR_Typ, Loc)),
+
+           Make_Object_Declaration (Loc,
+             Defining_Identifier => Rnn,
+             Object_Definition   => New_Occurrence_Of (QR_Typ, Loc)),
+
+           Make_Procedure_Call_Statement (Loc,
+             Name => New_Occurrence_Of (RTE (RE_Double_Divide), Loc),
+             Parameter_Associations => New_List (
+               Build_Conversion (N, QR_Typ, X),
+               Build_Conversion (N, QR_Typ, Y),
+               Build_Conversion (N, QR_Typ, Z),
+               New_Occurrence_Of (Qnn, Loc),
+               New_Occurrence_Of (Rnn, Loc),
+               New_Occurrence_Of (Rnd, Loc))));
+      end if;
+   end Build_Double_Divide_Code;
+
+   --------------------
+   -- Build_Multiply --
+   --------------------
+
+   function Build_Multiply (N : Node_Id; L, R : Node_Id) return Node_Id is
+      Loc         : constant Source_Ptr := Sloc (N);
+      Left_Type   : constant Entity_Id  := Etype (L);
+      Right_Type  : constant Entity_Id  := Etype (R);
+      Left_Size   : Int;
+      Right_Size  : Int;
+      Rsize       : Int;
+      Result_Type : Entity_Id;
+      Rnode       : Node_Id;
+
+   begin
+      --  Deal with floating-point case first
+
+      if Is_Floating_Point_Type (Left_Type) then
+         pragma Assert (Left_Type = Universal_Real);
+         pragma Assert (Right_Type = Universal_Real);
+
+         Result_Type := Universal_Real;
+         Rnode := Make_Op_Multiply (Loc, L, R);
+
+      --  Integer and fixed-point cases
+
+      else
+         --  An optimization. If the right operand is the literal 1, then we
+         --  can just return the left hand operand. Putting the optimization
+         --  here allows us to omit the check at the call site. Similarly, if
+         --  the left operand is the integer 1 we can return the right operand.
+
+         if Nkind (R) = N_Integer_Literal and then Intval (R) = 1 then
+            return L;
+         elsif Nkind (L) = N_Integer_Literal and then Intval (L) = 1 then
+            return R;
+         end if;
+
+         --  Otherwise we need to figure out the correct result type size
+         --  First figure out the effective sizes of the operands. Normally
+         --  the effective size of an operand is the RM_Size of the operand.
+         --  But a special case arises with operands whose size is known at
+         --  compile time. In this case, we can use the actual value of the
+         --  operand to get its size if it would fit signed in 8 or 16 bits.
+
+         Left_Size := UI_To_Int (RM_Size (Left_Type));
+
+         if Compile_Time_Known_Value (L) then
+            declare
+               Val : constant Uint := Expr_Value (L);
+            begin
+               if Val < Int'(2 ** 7) then
+                  Left_Size := 8;
+               elsif Val < Int'(2 ** 15) then
+                  Left_Size := 16;
+               end if;
+            end;
+         end if;
+
+         Right_Size := UI_To_Int (RM_Size (Right_Type));
+
+         if Compile_Time_Known_Value (R) then
+            declare
+               Val : constant Uint := Expr_Value (R);
+            begin
+               if Val <= Int'(2 ** 7) then
+                  Right_Size := 8;
+               elsif Val <= Int'(2 ** 15) then
+                  Right_Size := 16;
+               end if;
+            end;
+         end if;
+
+         --  Now the result size must be at least twice the longer of
+         --  the two sizes, to accommodate all possible results.
+
+         Rsize := 2 * Int'Max (Left_Size, Right_Size);
+
+         if Rsize <= 8 then
+            Result_Type := Standard_Integer_8;
+
+         elsif Rsize <= 16 then
+            Result_Type := Standard_Integer_16;
+
+         elsif Rsize <= 32 then
+            Result_Type := Standard_Integer_32;
+
+         else
+            Result_Type := Standard_Integer_64;
+         end if;
+
+         Rnode :=
+            Make_Op_Multiply (Loc,
+              Left_Opnd  => Build_Conversion (N, Result_Type, L),
+              Right_Opnd => Build_Conversion (N, Result_Type, R));
+      end if;
+
+      --  We now have a multiply node built with Result_Type set. First
+      --  set Etype of result, as required for all Build_xxx routines
+
+      Set_Etype (Rnode, Base_Type (Result_Type));
+
+      --  Set Treat_Fixed_As_Integer if operation on fixed-point type
+      --  since this is a literal arithmetic operation, to be performed
+      --  by Gigi without any consideration of small values.
+
+      if Is_Fixed_Point_Type (Result_Type) then
+         Set_Treat_Fixed_As_Integer (Rnode);
+      end if;
+
+      return Rnode;
+   end Build_Multiply;
+
+   ---------------
+   -- Build_Rem --
+   ---------------
+
+   function Build_Rem (N : Node_Id; L, R : Node_Id) return Node_Id is
+      Loc         : constant Source_Ptr := Sloc (N);
+      Left_Type   : constant Entity_Id  := Etype (L);
+      Right_Type  : constant Entity_Id  := Etype (R);
+      Result_Type : Entity_Id;
+      Rnode       : Node_Id;
+
+   begin
+      if Left_Type = Right_Type then
+         Result_Type := Left_Type;
+         Rnode :=
+           Make_Op_Rem (Loc,
+             Left_Opnd  => L,
+             Right_Opnd => R);
+
+      --  If left size is larger, we do the remainder operation using the
+      --  size of the left type (i.e. the larger of the two integer types).
+
+      elsif Esize (Left_Type) >= Esize (Right_Type) then
+         Result_Type := Left_Type;
+         Rnode :=
+           Make_Op_Rem (Loc,
+             Left_Opnd  => L,
+             Right_Opnd => Build_Conversion (N, Left_Type, R));
+
+      --  Similarly, if the right size is larger, we do the remainder
+      --  operation using the right type.
+
+      else
+         Result_Type := Right_Type;
+         Rnode :=
+           Make_Op_Rem (Loc,
+             Left_Opnd => Build_Conversion (N, Right_Type, L),
+             Right_Opnd => R);
+      end if;
+
+      --  We now have an N_Op_Rem node built with Result_Type set. First
+      --  set Etype of result, as required for all Build_xxx routines
+
+      Set_Etype (Rnode, Base_Type (Result_Type));
+
+      --  Set Treat_Fixed_As_Integer if operation on fixed-point type
+      --  since this is a literal arithmetic operation, to be performed
+      --  by Gigi without any consideration of small values.
+
+      if Is_Fixed_Point_Type (Result_Type) then
+         Set_Treat_Fixed_As_Integer (Rnode);
+      end if;
+
+      --  One more check. We did the rem operation using the larger of the
+      --  two types, which is reasonable. However, in the case where the
+      --  two types have unequal sizes, it is impossible for the result of
+      --  a remainder operation to be larger than the smaller of the two
+      --  types, so we can put a conversion round the result to keep the
+      --  evolving operation size as small as possible.
+
+      if Esize (Left_Type) >= Esize (Right_Type) then
+         Rnode := Build_Conversion (N, Right_Type, Rnode);
+      elsif Esize (Right_Type) >= Esize (Left_Type) then
+         Rnode := Build_Conversion (N, Left_Type, Rnode);
+      end if;
+
+      return Rnode;
+   end Build_Rem;
+
+   -------------------------
+   -- Build_Scaled_Divide --
+   -------------------------
+
+   function Build_Scaled_Divide
+     (N       : Node_Id;
+      X, Y, Z : Node_Id) return Node_Id
+   is
+      X_Size : constant Nat := UI_To_Int (Esize (Etype (X)));
+      Y_Size : constant Nat := UI_To_Int (Esize (Etype (Y)));
+      Expr   : Node_Id;
+
+   begin
+      --  If numerator fits in 64 bits, we can build the operations directly
+      --  without causing any intermediate overflow, so that's what we do.
+
+      if Nat'Max (X_Size, Y_Size) <= 32 then
+         return
+           Build_Divide (N, Build_Multiply (N, X, Y), Z);
+
+      --  Otherwise we use the runtime routine
+
+      --    [Qnn : Integer_64,
+      --     Rnn : Integer_64;
+      --     Scaled_Divide (X, Y, Z, Qnn, Rnn, Round);
+      --     Qnn]
+
+      else
+         declare
+            Loc  : constant Source_Ptr := Sloc (N);
+            Qnn  : Entity_Id;
+            Rnn  : Entity_Id;
+            Code : List_Id;
+
+            pragma Warnings (Off, Rnn);
+
+         begin
+            Build_Scaled_Divide_Code (N, X, Y, Z, Qnn, Rnn, Code);
+            Insert_Actions (N, Code);
+            Expr := New_Occurrence_Of (Qnn, Loc);
+
+            --  Set type of result in case used elsewhere (see note at start)
+
+            Set_Etype (Expr, Etype (Qnn));
+            return Expr;
+         end;
+      end if;
+   end Build_Scaled_Divide;
+
+   ------------------------------
+   -- Build_Scaled_Divide_Code --
+   ------------------------------
+
+   --  If the numerator can be computed in 64-bits, we build
+
+   --    [Nnn : constant typ := typ (X) * typ (Y);
+   --     Dnn : constant typ := typ (Z)
+   --     Qnn : constant typ := Nnn / Dnn;
+   --     Rnn : constant typ := Nnn / Dnn;
+
+   --  If the numerator cannot be computed in 64 bits, we build
+
+   --    [Qnn : Interfaces.Integer_64;
+   --     Rnn : Interfaces.Integer_64;
+   --     Scaled_Divide (X, Y, Z, Qnn, Rnn, Round);]
+
+   procedure Build_Scaled_Divide_Code
+     (N        : Node_Id;
+      X, Y, Z  : Node_Id;
+      Qnn, Rnn : out Entity_Id;
+      Code     : out List_Id)
+   is
+      Loc    : constant Source_Ptr := Sloc (N);
+
+      X_Size : constant Nat := UI_To_Int (Esize (Etype (X)));
+      Y_Size : constant Nat := UI_To_Int (Esize (Etype (Y)));
+      Z_Size : constant Nat := UI_To_Int (Esize (Etype (Z)));
+
+      QR_Siz : Nat;
+      QR_Typ : Entity_Id;
+
+      Nnn : Entity_Id;
+      Dnn : Entity_Id;
+
+      Quo : Node_Id;
+      Rnd : Entity_Id;
+
+   begin
+      --  Find type that will allow computation of numerator
+
+      QR_Siz := Nat'Max (X_Size, 2 * Nat'Max (Y_Size, Z_Size));
+
+      if QR_Siz <= 16 then
+         QR_Typ := Standard_Integer_16;
+      elsif QR_Siz <= 32 then
+         QR_Typ := Standard_Integer_32;
+      elsif QR_Siz <= 64 then
+         QR_Typ := Standard_Integer_64;
+
+      --  For more than 64, bits, we use the 64-bit integer defined in
+      --  Interfaces, so that it can be handled by the runtime routine.
+
+      else
+         QR_Typ := RTE (RE_Integer_64);
+      end if;
+
+      --  Define quotient and remainder, and set their Etypes, so
+      --  that they can be picked up by Build_xxx routines.
+
+      Qnn := Make_Temporary (Loc, 'S');
+      Rnn := Make_Temporary (Loc, 'R');
+
+      Set_Etype (Qnn, QR_Typ);
+      Set_Etype (Rnn, QR_Typ);
+
+      --  Case that we can compute the numerator in 64 bits
+
+      if QR_Siz <= 64 then
+         Nnn := Make_Temporary (Loc, 'N');
+         Dnn := Make_Temporary (Loc, 'D');
+
+         --  Set Etypes, so that they can be picked up by New_Occurrence_Of
+
+         Set_Etype (Nnn, QR_Typ);
+         Set_Etype (Dnn, QR_Typ);
+
+         Code := New_List (
+           Make_Object_Declaration (Loc,
+             Defining_Identifier => Nnn,
+             Object_Definition   => New_Occurrence_Of (QR_Typ, Loc),
+             Constant_Present    => True,
+             Expression =>
+               Build_Multiply (N,
+                 Build_Conversion (N, QR_Typ, X),
+                 Build_Conversion (N, QR_Typ, Y))),
+
+           Make_Object_Declaration (Loc,
+             Defining_Identifier => Dnn,
+             Object_Definition   => New_Occurrence_Of (QR_Typ, Loc),
+             Constant_Present    => True,
+             Expression => Build_Conversion (N, QR_Typ, Z)));
+
+         Quo :=
+           Build_Divide (N,
+             New_Occurrence_Of (Nnn, Loc),
+             New_Occurrence_Of (Dnn, Loc));
+
+         Append_To (Code,
+           Make_Object_Declaration (Loc,
+             Defining_Identifier => Qnn,
+             Object_Definition   => New_Occurrence_Of (QR_Typ, Loc),
+             Constant_Present    => True,
+             Expression          => Quo));
+
+         Append_To (Code,
+           Make_Object_Declaration (Loc,
+             Defining_Identifier => Rnn,
+             Object_Definition   => New_Occurrence_Of (QR_Typ, Loc),
+             Constant_Present    => True,
+             Expression =>
+               Build_Rem (N,
+                 New_Occurrence_Of (Nnn, Loc),
+                 New_Occurrence_Of (Dnn, Loc))));
+
+      --  Case where numerator does not fit in 64 bits, so we have to
+      --  call the runtime routine to compute the quotient and remainder
+
+      else
+         Rnd := Boolean_Literals (Rounded_Result_Set (N));
+
+         Code := New_List (
+           Make_Object_Declaration (Loc,
+             Defining_Identifier => Qnn,
+             Object_Definition   => New_Occurrence_Of (QR_Typ, Loc)),
+
+           Make_Object_Declaration (Loc,
+             Defining_Identifier => Rnn,
+             Object_Definition   => New_Occurrence_Of (QR_Typ, Loc)),
+
+           Make_Procedure_Call_Statement (Loc,
+             Name => New_Occurrence_Of (RTE (RE_Scaled_Divide), Loc),
+             Parameter_Associations => New_List (
+               Build_Conversion (N, QR_Typ, X),
+               Build_Conversion (N, QR_Typ, Y),
+               Build_Conversion (N, QR_Typ, Z),
+               New_Occurrence_Of (Qnn, Loc),
+               New_Occurrence_Of (Rnn, Loc),
+               New_Occurrence_Of (Rnd, Loc))));
+      end if;
+
+      --  Set type of result, for use in caller
+
+      Set_Etype (Qnn, QR_Typ);
+   end Build_Scaled_Divide_Code;
+
+   ---------------------------
+   -- Do_Divide_Fixed_Fixed --
+   ---------------------------
+
+   --  We have:
+
+   --    (Result_Value * Result_Small) =
+   --        (Left_Value * Left_Small) / (Right_Value * Right_Small)
+
+   --    Result_Value = (Left_Value / Right_Value) *
+   --                   (Left_Small / (Right_Small * Result_Small));
+
+   --  we can do the operation in integer arithmetic if this fraction is an
+   --  integer or the reciprocal of an integer, as detailed in (RM G.2.3(21)).
+   --  Otherwise the result is in the close result set and our approach is to
+   --  use floating-point to compute this close result.
+
+   procedure Do_Divide_Fixed_Fixed (N : Node_Id) is
+      Left        : constant Node_Id   := Left_Opnd (N);
+      Right       : constant Node_Id   := Right_Opnd (N);
+      Left_Type   : constant Entity_Id := Etype (Left);
+      Right_Type  : constant Entity_Id := Etype (Right);
+      Result_Type : constant Entity_Id := Etype (N);
+      Right_Small : constant Ureal     := Small_Value (Right_Type);
+      Left_Small  : constant Ureal     := Small_Value (Left_Type);
+
+      Result_Small : Ureal;
+      Frac         : Ureal;
+      Frac_Num     : Uint;
+      Frac_Den     : Uint;
+      Lit_Int      : Node_Id;
+
+   begin
+      --  Rounding is required if the result is integral
+
+      if Is_Integer_Type (Result_Type) then
+         Set_Rounded_Result (N);
+      end if;
+
+      --  Get result small. If the result is an integer, treat it as though
+      --  it had a small of 1.0, all other processing is identical.
+
+      if Is_Integer_Type (Result_Type) then
+         Result_Small := Ureal_1;
+      else
+         Result_Small := Small_Value (Result_Type);
+      end if;
+
+      --  Get small ratio
+
+      Frac     := Left_Small / (Right_Small * Result_Small);
+      Frac_Num := Norm_Num (Frac);
+      Frac_Den := Norm_Den (Frac);
+
+      --  If the fraction is an integer, then we get the result by multiplying
+      --  the left operand by the integer, and then dividing by the right
+      --  operand (the order is important, if we did the divide first, we
+      --  would lose precision).
+
+      if Frac_Den = 1 then
+         Lit_Int := Integer_Literal (N, Frac_Num); -- always positive
+
+         if Present (Lit_Int) then
+            Set_Result (N, Build_Scaled_Divide (N, Left, Lit_Int, Right));
+            return;
+         end if;
+
+      --  If the fraction is the reciprocal of an integer, then we get the
+      --  result by first multiplying the divisor by the integer, and then
+      --  doing the division with the adjusted divisor.
+
+      --  Note: this is much better than doing two divisions: multiplications
+      --  are much faster than divisions (and certainly faster than rounded
+      --  divisions), and we don't get inaccuracies from double rounding.
+
+      elsif Frac_Num = 1 then
+         Lit_Int := Integer_Literal (N, Frac_Den); -- always positive
+
+         if Present (Lit_Int) then
+            Set_Result (N, Build_Double_Divide (N, Left, Right, Lit_Int));
+            return;
+         end if;
+      end if;
+
+      --  If we fall through, we use floating-point to compute the result
+
+      Set_Result (N,
+        Build_Multiply (N,
+          Build_Divide (N, Fpt_Value (Left), Fpt_Value (Right)),
+          Real_Literal (N, Frac)));
+   end Do_Divide_Fixed_Fixed;
+
+   -------------------------------
+   -- Do_Divide_Fixed_Universal --
+   -------------------------------
+
+   --  We have:
+
+   --    (Result_Value * Result_Small) = (Left_Value * Left_Small) / Lit_Value;
+   --    Result_Value = Left_Value * Left_Small /(Lit_Value * Result_Small);
+
+   --  The result is required to be in the perfect result set if the literal
+   --  can be factored so that the resulting small ratio is an integer or the
+   --  reciprocal of an integer (RM G.2.3(21-22)). We now give a detailed
+   --  analysis of these RM requirements:
+
+   --  We must factor the literal, finding an integer K:
+
+   --     Lit_Value = K * Right_Small
+   --     Right_Small = Lit_Value / K
+
+   --  such that the small ratio:
+
+   --              Left_Small
+   --     ------------------------------
+   --     (Lit_Value / K) * Result_Small
+
+   --            Left_Small
+   --  =  ------------------------  *  K
+   --     Lit_Value * Result_Small
+
+   --  is an integer or the reciprocal of an integer, and for
+   --  implementation efficiency we need the smallest such K.
+
+   --  First we reduce the left fraction to lowest terms
+
+   --    If numerator = 1, then for K = 1, the small ratio is the reciprocal
+   --    of an integer, and this is clearly the minimum K case, so set K = 1,
+   --    Right_Small = Lit_Value.
+
+   --    If numerator > 1, then set K to the denominator of the fraction so
+   --    that the resulting small ratio is an integer (the numerator value).
+
+   procedure Do_Divide_Fixed_Universal (N : Node_Id) is
+      Left        : constant Node_Id   := Left_Opnd (N);
+      Right       : constant Node_Id   := Right_Opnd (N);
+      Left_Type   : constant Entity_Id := Etype (Left);
+      Result_Type : constant Entity_Id := Etype (N);
+      Left_Small  : constant Ureal     := Small_Value (Left_Type);
+      Lit_Value   : constant Ureal     := Realval (Right);
+
+      Result_Small : Ureal;
+      Frac         : Ureal;
+      Frac_Num     : Uint;
+      Frac_Den     : Uint;
+      Lit_K        : Node_Id;
+      Lit_Int      : Node_Id;
+
+   begin
+      --  Get result small. If the result is an integer, treat it as though
+      --  it had a small of 1.0, all other processing is identical.
+
+      if Is_Integer_Type (Result_Type) then
+         Result_Small := Ureal_1;
+      else
+         Result_Small := Small_Value (Result_Type);
+      end if;
+
+      --  Determine if literal can be rewritten successfully
+
+      Frac     := Left_Small / (Lit_Value * Result_Small);
+      Frac_Num := Norm_Num (Frac);
+      Frac_Den := Norm_Den (Frac);
+
+      --  Case where fraction is the reciprocal of an integer (K = 1, integer
+      --  = denominator). If this integer is not too large, this is the case
+      --  where the result can be obtained by dividing by this integer value.
+
+      if Frac_Num = 1 then
+         Lit_Int := Integer_Literal (N, Frac_Den, UR_Is_Negative (Frac));
+
+         if Present (Lit_Int) then
+            Set_Result (N, Build_Divide (N, Left, Lit_Int));
+            return;
+         end if;
+
+      --  Case where we choose K to make fraction an integer (K = denominator
+      --  of fraction, integer = numerator of fraction). If both K and the
+      --  numerator are small enough, this is the case where the result can
+      --  be obtained by first multiplying by the integer value and then
+      --  dividing by K (the order is important, if we divided first, we
+      --  would lose precision).
+
+      else
+         Lit_Int := Integer_Literal (N, Frac_Num, UR_Is_Negative (Frac));
+         Lit_K   := Integer_Literal (N, Frac_Den, False);
+
+         if Present (Lit_Int) and then Present (Lit_K) then
+            Set_Result (N, Build_Scaled_Divide (N, Left, Lit_Int, Lit_K));
+            return;
+         end if;
+      end if;
+
+      --  Fall through if the literal cannot be successfully rewritten, or if
+      --  the small ratio is out of range of integer arithmetic. In the former
+      --  case it is fine to use floating-point to get the close result set,
+      --  and in the latter case, it means that the result is zero or raises
+      --  constraint error, and we can do that accurately in floating-point.
+
+      --  If we end up using floating-point, then we take the right integer
+      --  to be one, and its small to be the value of the original right real
+      --  literal. That way, we need only one floating-point multiplication.
+
+      Set_Result (N,
+        Build_Multiply (N, Fpt_Value (Left), Real_Literal (N, Frac)));
+   end Do_Divide_Fixed_Universal;
+
+   -------------------------------
+   -- Do_Divide_Universal_Fixed --
+   -------------------------------
+
+   --  We have:
+
+   --    (Result_Value * Result_Small) =
+   --          Lit_Value / (Right_Value * Right_Small)
+   --    Result_Value =
+   --          (Lit_Value / (Right_Small * Result_Small)) / Right_Value
+
+   --  The result is required to be in the perfect result set if the literal
+   --  can be factored so that the resulting small ratio is an integer or the
+   --  reciprocal of an integer (RM G.2.3(21-22)). We now give a detailed
+   --  analysis of these RM requirements:
+
+   --  We must factor the literal, finding an integer K:
+
+   --     Lit_Value = K * Left_Small
+   --     Left_Small = Lit_Value / K
+
+   --  such that the small ratio:
+
+   --           (Lit_Value / K)
+   --     --------------------------
+   --     Right_Small * Result_Small
+
+   --              Lit_Value             1
+   --  =  --------------------------  *  -
+   --     Right_Small * Result_Small     K
+
+   --  is an integer or the reciprocal of an integer, and for
+   --  implementation efficiency we need the smallest such K.
+
+   --  First we reduce the left fraction to lowest terms
+
+   --    If denominator = 1, then for K = 1, the small ratio is an integer
+   --    (the numerator) and this is clearly the minimum K case, so set K = 1,
+   --    and Left_Small = Lit_Value.
+
+   --    If denominator > 1, then set K to the numerator of the fraction so
+   --    that the resulting small ratio is the reciprocal of an integer (the
+   --    numerator value).
+
+   procedure Do_Divide_Universal_Fixed (N : Node_Id) is
+      Left        : constant Node_Id   := Left_Opnd (N);
+      Right       : constant Node_Id   := Right_Opnd (N);
+      Right_Type  : constant Entity_Id := Etype (Right);
+      Result_Type : constant Entity_Id := Etype (N);
+      Right_Small : constant Ureal     := Small_Value (Right_Type);
+      Lit_Value   : constant Ureal     := Realval (Left);
+
+      Result_Small : Ureal;
+      Frac         : Ureal;
+      Frac_Num     : Uint;
+      Frac_Den     : Uint;
+      Lit_K        : Node_Id;
+      Lit_Int      : Node_Id;
+
+   begin
+      --  Get result small. If the result is an integer, treat it as though
+      --  it had a small of 1.0, all other processing is identical.
+
+      if Is_Integer_Type (Result_Type) then
+         Result_Small := Ureal_1;
+      else
+         Result_Small := Small_Value (Result_Type);
+      end if;
+
+      --  Determine if literal can be rewritten successfully
+
+      Frac     := Lit_Value / (Right_Small * Result_Small);
+      Frac_Num := Norm_Num (Frac);
+      Frac_Den := Norm_Den (Frac);
+
+      --  Case where fraction is an integer (K = 1, integer = numerator). If
+      --  this integer is not too large, this is the case where the result
+      --  can be obtained by dividing this integer by the right operand.
+
+      if Frac_Den = 1 then
+         Lit_Int := Integer_Literal (N, Frac_Num, UR_Is_Negative (Frac));
+
+         if Present (Lit_Int) then
+            Set_Result (N, Build_Divide (N, Lit_Int, Right));
+            return;
+         end if;
+
+      --  Case where we choose K to make the fraction the reciprocal of an
+      --  integer (K = numerator of fraction, integer = numerator of fraction).
+      --  If both K and the integer are small enough, this is the case where
+      --  the result can be obtained by multiplying the right operand by K
+      --  and then dividing by the integer value. The order of the operations
+      --  is important (if we divided first, we would lose precision).
+
+      else
+         Lit_Int := Integer_Literal (N, Frac_Den, UR_Is_Negative (Frac));
+         Lit_K   := Integer_Literal (N, Frac_Num, False);
+
+         if Present (Lit_Int) and then Present (Lit_K) then
+            Set_Result (N, Build_Double_Divide (N, Lit_K, Right, Lit_Int));
+            return;
+         end if;
+      end if;
+
+      --  Fall through if the literal cannot be successfully rewritten, or if
+      --  the small ratio is out of range of integer arithmetic. In the former
+      --  case it is fine to use floating-point to get the close result set,
+      --  and in the latter case, it means that the result is zero or raises
+      --  constraint error, and we can do that accurately in floating-point.
+
+      --  If we end up using floating-point, then we take the right integer
+      --  to be one, and its small to be the value of the original right real
+      --  literal. That way, we need only one floating-point division.
+
+      Set_Result (N,
+        Build_Divide (N, Real_Literal (N, Frac), Fpt_Value (Right)));
+   end Do_Divide_Universal_Fixed;
+
+   -----------------------------
+   -- Do_Multiply_Fixed_Fixed --
+   -----------------------------
+
+   --  We have:
+
+   --    (Result_Value * Result_Small) =
+   --        (Left_Value * Left_Small) * (Right_Value * Right_Small)
+
+   --    Result_Value = (Left_Value * Right_Value) *
+   --                   (Left_Small * Right_Small) / Result_Small;
+
+   --  we can do the operation in integer arithmetic if this fraction is an
+   --  integer or the reciprocal of an integer, as detailed in (RM G.2.3(21)).
+   --  Otherwise the result is in the close result set and our approach is to
+   --  use floating-point to compute this close result.
+
+   procedure Do_Multiply_Fixed_Fixed (N : Node_Id) is
+      Left  : constant Node_Id := Left_Opnd (N);
+      Right : constant Node_Id := Right_Opnd (N);
+
+      Left_Type   : constant Entity_Id := Etype (Left);
+      Right_Type  : constant Entity_Id := Etype (Right);
+      Result_Type : constant Entity_Id := Etype (N);
+      Right_Small : constant Ureal     := Small_Value (Right_Type);
+      Left_Small  : constant Ureal     := Small_Value (Left_Type);
+
+      Result_Small : Ureal;
+      Frac         : Ureal;
+      Frac_Num     : Uint;
+      Frac_Den     : Uint;
+      Lit_Int      : Node_Id;
+
+   begin
+      --  Get result small. If the result is an integer, treat it as though
+      --  it had a small of 1.0, all other processing is identical.
+
+      if Is_Integer_Type (Result_Type) then
+         Result_Small := Ureal_1;
+      else
+         Result_Small := Small_Value (Result_Type);
+      end if;
+
+      --  Get small ratio
+
+      Frac     := (Left_Small * Right_Small) / Result_Small;
+      Frac_Num := Norm_Num (Frac);
+      Frac_Den := Norm_Den (Frac);
+
+      --  If the fraction is an integer, then we get the result by multiplying
+      --  the operands, and then multiplying the result by the integer value.
+
+      if Frac_Den = 1 then
+         Lit_Int := Integer_Literal (N, Frac_Num); -- always positive
+
+         if Present (Lit_Int) then
+            Set_Result (N,
+              Build_Multiply (N, Build_Multiply (N, Left, Right),
+                Lit_Int));
+            return;
+         end if;
+
+      --  If the fraction is the reciprocal of an integer, then we get the
+      --  result by multiplying the operands, and then dividing the result by
+      --  the integer value. The order of the operations is important, if we
+      --  divided first, we would lose precision.
+
+      elsif Frac_Num = 1 then
+         Lit_Int := Integer_Literal (N, Frac_Den); -- always positive
+
+         if Present (Lit_Int) then
+            Set_Result (N, Build_Scaled_Divide (N, Left, Right, Lit_Int));
+            return;
+         end if;
+      end if;
+
+      --  If we fall through, we use floating-point to compute the result
+
+      Set_Result (N,
+        Build_Multiply (N,
+          Build_Multiply (N, Fpt_Value (Left), Fpt_Value (Right)),
+          Real_Literal (N, Frac)));
+   end Do_Multiply_Fixed_Fixed;
+
+   ---------------------------------
+   -- Do_Multiply_Fixed_Universal --
+   ---------------------------------
+
+   --  We have:
+
+   --    (Result_Value * Result_Small) = (Left_Value * Left_Small) * Lit_Value;
+   --    Result_Value = Left_Value * (Left_Small * Lit_Value) / Result_Small;
+
+   --  The result is required to be in the perfect result set if the literal
+   --  can be factored so that the resulting small ratio is an integer or the
+   --  reciprocal of an integer (RM G.2.3(21-22)). We now give a detailed
+   --  analysis of these RM requirements:
+
+   --  We must factor the literal, finding an integer K:
+
+   --     Lit_Value = K * Right_Small
+   --     Right_Small = Lit_Value / K
+
+   --  such that the small ratio:
+
+   --     Left_Small * (Lit_Value / K)
+   --     ----------------------------
+   --             Result_Small
+
+   --     Left_Small * Lit_Value     1
+   --  =  ----------------------  *  -
+   --          Result_Small          K
+
+   --  is an integer or the reciprocal of an integer, and for
+   --  implementation efficiency we need the smallest such K.
+
+   --  First we reduce the left fraction to lowest terms
+
+   --    If denominator = 1, then for K = 1, the small ratio is an integer, and
+   --    this is clearly the minimum K case, so set
+
+   --      K = 1, Right_Small = Lit_Value
+
+   --    If denominator > 1, then set K to the numerator of the fraction, so
+   --    that the resulting small ratio is the reciprocal of the integer (the
+   --    denominator value).
+
+   procedure Do_Multiply_Fixed_Universal
+     (N           : Node_Id;
+      Left, Right : Node_Id)
+   is
+      Left_Type   : constant Entity_Id := Etype (Left);
+      Result_Type : constant Entity_Id := Etype (N);
+      Left_Small  : constant Ureal     := Small_Value (Left_Type);
+      Lit_Value   : constant Ureal     := Realval (Right);
+
+      Result_Small : Ureal;
+      Frac         : Ureal;
+      Frac_Num     : Uint;
+      Frac_Den     : Uint;
+      Lit_K        : Node_Id;
+      Lit_Int      : Node_Id;
+
+   begin
+      --  Get result small. If the result is an integer, treat it as though
+      --  it had a small of 1.0, all other processing is identical.
+
+      if Is_Integer_Type (Result_Type) then
+         Result_Small := Ureal_1;
+      else
+         Result_Small := Small_Value (Result_Type);
+      end if;
+
+      --  Determine if literal can be rewritten successfully
+
+      Frac     := (Left_Small * Lit_Value) / Result_Small;
+      Frac_Num := Norm_Num (Frac);
+      Frac_Den := Norm_Den (Frac);
+
+      --  Case where fraction is an integer (K = 1, integer = numerator). If
+      --  this integer is not too large, this is the case where the result can
+      --  be obtained by multiplying by this integer value.
+
+      if Frac_Den = 1 then
+         Lit_Int := Integer_Literal (N, Frac_Num, UR_Is_Negative (Frac));
+
+         if Present (Lit_Int) then
+            Set_Result (N, Build_Multiply (N, Left, Lit_Int));
+            return;
+         end if;
+
+      --  Case where we choose K to make fraction the reciprocal of an integer
+      --  (K = numerator of fraction, integer = denominator of fraction). If
+      --  both K and the denominator are small enough, this is the case where
+      --  the result can be obtained by first multiplying by K, and then
+      --  dividing by the integer value.
+
+      else
+         Lit_Int := Integer_Literal (N, Frac_Den, UR_Is_Negative (Frac));
+         Lit_K   := Integer_Literal (N, Frac_Num);
+
+         if Present (Lit_Int) and then Present (Lit_K) then
+            Set_Result (N, Build_Scaled_Divide (N, Left, Lit_K, Lit_Int));
+            return;
+         end if;
+      end if;
+
+      --  Fall through if the literal cannot be successfully rewritten, or if
+      --  the small ratio is out of range of integer arithmetic. In the former
+      --  case it is fine to use floating-point to get the close result set,
+      --  and in the latter case, it means that the result is zero or raises
+      --  constraint error, and we can do that accurately in floating-point.
+
+      --  If we end up using floating-point, then we take the right integer
+      --  to be one, and its small to be the value of the original right real
+      --  literal. That way, we need only one floating-point multiplication.
+
+      Set_Result (N,
+        Build_Multiply (N, Fpt_Value (Left), Real_Literal (N, Frac)));
+   end Do_Multiply_Fixed_Universal;
+
+   ---------------------------------
+   -- Expand_Convert_Fixed_Static --
+   ---------------------------------
+
+   procedure Expand_Convert_Fixed_Static (N : Node_Id) is
+   begin
+      Rewrite (N,
+        Convert_To (Etype (N),
+          Make_Real_Literal (Sloc (N), Expr_Value_R (Expression (N)))));
+      Analyze_And_Resolve (N);
+   end Expand_Convert_Fixed_Static;
+
+   -----------------------------------
+   -- Expand_Convert_Fixed_To_Fixed --
+   -----------------------------------
+
+   --  We have:
+
+   --    Result_Value * Result_Small = Source_Value * Source_Small
+   --    Result_Value = Source_Value * (Source_Small / Result_Small)
+
+   --  If the small ratio (Source_Small / Result_Small) is a sufficiently small
+   --  integer, then the perfect result set is obtained by a single integer
+   --  multiplication.
+
+   --  If the small ratio is the reciprocal of a sufficiently small integer,
+   --  then the perfect result set is obtained by a single integer division.
+
+   --  In other cases, we obtain the close result set by calculating the
+   --  result in floating-point.
+
+   procedure Expand_Convert_Fixed_To_Fixed (N : Node_Id) is
+      Rng_Check   : constant Boolean   := Do_Range_Check (N);
+      Expr        : constant Node_Id   := Expression (N);
+      Result_Type : constant Entity_Id := Etype (N);
+      Source_Type : constant Entity_Id := Etype (Expr);
+      Small_Ratio : Ureal;
+      Ratio_Num   : Uint;
+      Ratio_Den   : Uint;
+      Lit         : Node_Id;
+
+   begin
+      if Is_OK_Static_Expression (Expr) then
+         Expand_Convert_Fixed_Static (N);
+         return;
+      end if;
+
+      Small_Ratio := Small_Value (Source_Type) / Small_Value (Result_Type);
+      Ratio_Num   := Norm_Num (Small_Ratio);
+      Ratio_Den   := Norm_Den (Small_Ratio);
+
+      if Ratio_Den = 1 then
+         if Ratio_Num = 1 then
+            Set_Result (N, Expr);
+            return;
+
+         else
+            Lit := Integer_Literal (N, Ratio_Num);
+
+            if Present (Lit) then
+               Set_Result (N, Build_Multiply (N, Expr, Lit));
+               return;
+            end if;
+         end if;
+
+      elsif Ratio_Num = 1 then
+         Lit := Integer_Literal (N, Ratio_Den);
+
+         if Present (Lit) then
+            Set_Result (N, Build_Divide (N, Expr, Lit), Rng_Check);
+            return;
+         end if;
+      end if;
+
+      --  Fall through to use floating-point for the close result set case
+      --  either as a result of the small ratio not being an integer or the
+      --  reciprocal of an integer, or if the integer is out of range.
+
+      Set_Result (N,
+        Build_Multiply (N,
+          Fpt_Value (Expr),
+          Real_Literal (N, Small_Ratio)),
+        Rng_Check);
+   end Expand_Convert_Fixed_To_Fixed;
+
+   -----------------------------------
+   -- Expand_Convert_Fixed_To_Float --
+   -----------------------------------
+
+   --  If the small of the fixed type is 1.0, then we simply convert the
+   --  integer value directly to the target floating-point type, otherwise
+   --  we first have to multiply by the small, in Universal_Real, and then
+   --  convert the result to the target floating-point type.
+
+   procedure Expand_Convert_Fixed_To_Float (N : Node_Id) is
+      Rng_Check   : constant Boolean    := Do_Range_Check (N);
+      Expr        : constant Node_Id    := Expression (N);
+      Source_Type : constant Entity_Id  := Etype (Expr);
+      Small       : constant Ureal      := Small_Value (Source_Type);
+
+   begin
+      if Is_OK_Static_Expression (Expr) then
+         Expand_Convert_Fixed_Static (N);
+         return;
+      end if;
+
+      if Small = Ureal_1 then
+         Set_Result (N, Expr);
+
+      else
+         Set_Result (N,
+           Build_Multiply (N,
+             Fpt_Value (Expr),
+             Real_Literal (N, Small)),
+           Rng_Check);
+      end if;
+   end Expand_Convert_Fixed_To_Float;
+
+   -------------------------------------
+   -- Expand_Convert_Fixed_To_Integer --
+   -------------------------------------
+
+   --  We have:
+
+   --    Result_Value = Source_Value * Source_Small
+
+   --  If the small value is a sufficiently small integer, then the perfect
+   --  result set is obtained by a single integer multiplication.
+
+   --  If the small value is the reciprocal of a sufficiently small integer,
+   --  then the perfect result set is obtained by a single integer division.
+
+   --  In other cases, we obtain the close result set by calculating the
+   --  result in floating-point.
+
+   procedure Expand_Convert_Fixed_To_Integer (N : Node_Id) is
+      Rng_Check   : constant Boolean   := Do_Range_Check (N);
+      Expr        : constant Node_Id   := Expression (N);
+      Source_Type : constant Entity_Id := Etype (Expr);
+      Small       : constant Ureal     := Small_Value (Source_Type);
+      Small_Num   : constant Uint      := Norm_Num (Small);
+      Small_Den   : constant Uint      := Norm_Den (Small);
+      Lit         : Node_Id;
+
+   begin
+      if Is_OK_Static_Expression (Expr) then
+         Expand_Convert_Fixed_Static (N);
+         return;
+      end if;
+
+      if Small_Den = 1 then
+         Lit := Integer_Literal (N, Small_Num);
+
+         if Present (Lit) then
+            Set_Result (N, Build_Multiply (N, Expr, Lit), Rng_Check);
+            return;
+         end if;
+
+      elsif Small_Num = 1 then
+         Lit := Integer_Literal (N, Small_Den);
+
+         if Present (Lit) then
+            Set_Result (N, Build_Divide (N, Expr, Lit), Rng_Check);
+            return;
+         end if;
+      end if;
+
+      --  Fall through to use floating-point for the close result set case
+      --  either as a result of the small value not being an integer or the
+      --  reciprocal of an integer, or if the integer is out of range.
+
+      Set_Result (N,
+        Build_Multiply (N,
+          Fpt_Value (Expr),
+          Real_Literal (N, Small)),
+        Rng_Check);
+   end Expand_Convert_Fixed_To_Integer;
+
+   -----------------------------------
+   -- Expand_Convert_Float_To_Fixed --
+   -----------------------------------
+
+   --  We have
+
+   --    Result_Value * Result_Small = Operand_Value
+
+   --  so compute:
+
+   --    Result_Value = Operand_Value * (1.0 / Result_Small)
+
+   --  We do the small scaling in floating-point, and we do a multiplication
+   --  rather than a division, since it is accurate enough for the perfect
+   --  result cases, and faster.
+
+   procedure Expand_Convert_Float_To_Fixed (N : Node_Id) is
+      Expr        : constant Node_Id   := Expression (N);
+      Orig_N      : constant Node_Id   := Original_Node (N);
+      Result_Type : constant Entity_Id := Etype (N);
+      Rng_Check   : constant Boolean   := Do_Range_Check (N);
+      Small       : constant Ureal     := Small_Value (Result_Type);
+      Truncate    : Boolean;
+
+   begin
+      --  Optimize small = 1, where we can avoid the multiply completely
+
+      if Small = Ureal_1 then
+         Set_Result (N, Expr, Rng_Check, Trunc => True);
+
+      --  Normal case where multiply is required. Rounding is truncating
+      --  for decimal fixed point types only, see RM 4.6(29), except if the
+      --  conversion comes from an attribute reference 'Round (RM 3.5.10 (14)):
+      --  The attribute is implemented by means of a conversion that must
+      --  round.
+
+      else
+         if Is_Decimal_Fixed_Point_Type (Result_Type) then
+            Truncate :=
+              Nkind (Orig_N) /= N_Attribute_Reference
+                or else Get_Attribute_Id
+                          (Attribute_Name (Orig_N)) /= Attribute_Round;
+         else
+            Truncate := False;
+         end if;
+
+         Set_Result
+           (N     => N,
+            Expr  =>
+              Build_Multiply
+                (N => N,
+                 L => Fpt_Value (Expr),
+                 R => Real_Literal (N, Ureal_1 / Small)),
+            Rchk  => Rng_Check,
+            Trunc => Truncate);
+      end if;
+   end Expand_Convert_Float_To_Fixed;
+
+   -------------------------------------
+   -- Expand_Convert_Integer_To_Fixed --
+   -------------------------------------
+
+   --  We have
+
+   --    Result_Value * Result_Small = Operand_Value
+   --    Result_Value = Operand_Value / Result_Small
+
+   --  If the small value is a sufficiently small integer, then the perfect
+   --  result set is obtained by a single integer division.
+
+   --  If the small value is the reciprocal of a sufficiently small integer,
+   --  the perfect result set is obtained by a single integer multiplication.
+
+   --  In other cases, we obtain the close result set by calculating the
+   --  result in floating-point using a multiplication by the reciprocal
+   --  of the Result_Small.
+
+   procedure Expand_Convert_Integer_To_Fixed (N : Node_Id) is
+      Rng_Check   : constant Boolean   := Do_Range_Check (N);
+      Expr        : constant Node_Id   := Expression (N);
+      Result_Type : constant Entity_Id := Etype (N);
+      Small       : constant Ureal     := Small_Value (Result_Type);
+      Small_Num   : constant Uint      := Norm_Num (Small);
+      Small_Den   : constant Uint      := Norm_Den (Small);
+      Lit         : Node_Id;
+
+   begin
+      if Small_Den = 1 then
+         Lit := Integer_Literal (N, Small_Num);
+
+         if Present (Lit) then
+            Set_Result (N, Build_Divide (N, Expr, Lit), Rng_Check);
+            return;
+         end if;
+
+      elsif Small_Num = 1 then
+         Lit := Integer_Literal (N, Small_Den);
+
+         if Present (Lit) then
+            Set_Result (N, Build_Multiply (N, Expr, Lit), Rng_Check);
+            return;
+         end if;
+      end if;
+
+      --  Fall through to use floating-point for the close result set case
+      --  either as a result of the small value not being an integer or the
+      --  reciprocal of an integer, or if the integer is out of range.
+
+      Set_Result (N,
+        Build_Multiply (N,
+          Fpt_Value (Expr),
+          Real_Literal (N, Ureal_1 / Small)),
+        Rng_Check);
+   end Expand_Convert_Integer_To_Fixed;
+
+   --------------------------------
+   -- Expand_Decimal_Divide_Call --
+   --------------------------------
+
+   --  We have four operands
+
+   --    Dividend
+   --    Divisor
+   --    Quotient
+   --    Remainder
+
+   --  All of which are decimal types, and which thus have associated
+   --  decimal scales.
+
+   --  Computing the quotient is a similar problem to that faced by the
+   --  normal fixed-point division, except that it is simpler, because
+   --  we always have compatible smalls.
+
+   --    Quotient = (Dividend / Divisor) * 10**q
+
+   --      where 10 ** q = Dividend'Small / (Divisor'Small * Quotient'Small)
+   --      so q = Divisor'Scale + Quotient'Scale - Dividend'Scale
+
+   --    For q >= 0, we compute
+
+   --      Numerator   := Dividend * 10 ** q
+   --      Denominator := Divisor
+   --      Quotient    := Numerator / Denominator
+
+   --    For q < 0, we compute
+
+   --      Numerator   := Dividend
+   --      Denominator := Divisor * 10 ** q
+   --      Quotient    := Numerator / Denominator
+
+   --  Both these divisions are done in truncated mode, and the remainder
+   --  from these divisions is used to compute the result Remainder. This
+   --  remainder has the effective scale of the numerator of the division,
+
+   --    For q >= 0, the remainder scale is Dividend'Scale + q
+   --    For q <  0, the remainder scale is Dividend'Scale
+
+   --  The result Remainder is then computed by a normal truncating decimal
+   --  conversion from this scale to the scale of the remainder, i.e. by a
+   --  division or multiplication by the appropriate power of 10.
+
+   procedure Expand_Decimal_Divide_Call (N : Node_Id) is
+      Loc : constant Source_Ptr := Sloc (N);
+
+      Dividend  : Node_Id := First_Actual (N);
+      Divisor   : Node_Id := Next_Actual (Dividend);
+      Quotient  : Node_Id := Next_Actual (Divisor);
+      Remainder : Node_Id := Next_Actual (Quotient);
+
+      Dividend_Type   : constant Entity_Id := Etype (Dividend);
+      Divisor_Type    : constant Entity_Id := Etype (Divisor);
+      Quotient_Type   : constant Entity_Id := Etype (Quotient);
+      Remainder_Type  : constant Entity_Id := Etype (Remainder);
+
+      Dividend_Scale  : constant Uint := Scale_Value (Dividend_Type);
+      Divisor_Scale   : constant Uint := Scale_Value (Divisor_Type);
+      Quotient_Scale  : constant Uint := Scale_Value (Quotient_Type);
+      Remainder_Scale : constant Uint := Scale_Value (Remainder_Type);
+
+      Q                  : Uint;
+      Numerator_Scale    : Uint;
+      Stmts              : List_Id;
+      Qnn                : Entity_Id;
+      Rnn                : Entity_Id;
+      Computed_Remainder : Node_Id;
+      Adjusted_Remainder : Node_Id;
+      Scale_Adjust       : Uint;
+
+   begin
+      --  Relocate the operands, since they are now list elements, and we
+      --  need to reference them separately as operands in the expanded code.
+
+      Dividend  := Relocate_Node (Dividend);
+      Divisor   := Relocate_Node (Divisor);
+      Quotient  := Relocate_Node (Quotient);
+      Remainder := Relocate_Node (Remainder);
+
+      --  Now compute Q, the adjustment scale
+
+      Q := Divisor_Scale + Quotient_Scale - Dividend_Scale;
+
+      --  If Q is non-negative then we need a scaled divide
+
+      if Q >= 0 then
+         Build_Scaled_Divide_Code
+           (N,
+            Dividend,
+            Integer_Literal (N, Uint_10 ** Q),
+            Divisor,
+            Qnn, Rnn, Stmts);
+
+         Numerator_Scale := Dividend_Scale + Q;
+
+      --  If Q is negative, then we need a double divide
+
+      else
+         Build_Double_Divide_Code
+           (N,
+            Dividend,
+            Divisor,
+            Integer_Literal (N, Uint_10 ** (-Q)),
+            Qnn, Rnn, Stmts);
+
+         Numerator_Scale := Dividend_Scale;
+      end if;
+
+      --  Add statement to set quotient value
+
+      --    Quotient := quotient-type!(Qnn);
+
+      Append_To (Stmts,
+        Make_Assignment_Statement (Loc,
+          Name => Quotient,
+          Expression =>
+            Unchecked_Convert_To (Quotient_Type,
+              Build_Conversion (N, Quotient_Type,
+                New_Occurrence_Of (Qnn, Loc)))));
+
+      --  Now we need to deal with computing and setting the remainder. The
+      --  scale of the remainder is in Numerator_Scale, and the desired
+      --  scale is the scale of the given Remainder argument. There are
+      --  three cases:
+
+      --    Numerator_Scale > Remainder_Scale
+
+      --      in this case, there are extra digits in the computed remainder
+      --      which must be eliminated by an extra division:
+
+      --        computed-remainder := Numerator rem Denominator
+      --        scale_adjust = Numerator_Scale - Remainder_Scale
+      --        adjusted-remainder := computed-remainder / 10 ** scale_adjust
+
+      --    Numerator_Scale = Remainder_Scale
+
+      --      in this case, the we have the remainder we need
+
+      --        computed-remainder := Numerator rem Denominator
+      --        adjusted-remainder := computed-remainder
+
+      --    Numerator_Scale < Remainder_Scale
+
+      --      in this case, we have insufficient digits in the computed
+      --      remainder, which must be eliminated by an extra multiply
+
+      --        computed-remainder := Numerator rem Denominator
+      --        scale_adjust = Remainder_Scale - Numerator_Scale
+      --        adjusted-remainder := computed-remainder * 10 ** scale_adjust
+
+      --  Finally we assign the adjusted-remainder to the result Remainder
+      --  with conversions to get the proper fixed-point type representation.
+
+      Computed_Remainder := New_Occurrence_Of (Rnn, Loc);
+
+      if Numerator_Scale > Remainder_Scale then
+         Scale_Adjust := Numerator_Scale - Remainder_Scale;
+         Adjusted_Remainder :=
+           Build_Divide
+             (N, Computed_Remainder, Integer_Literal (N, 10 ** Scale_Adjust));
+
+      elsif Numerator_Scale = Remainder_Scale then
+         Adjusted_Remainder := Computed_Remainder;
+
+      else -- Numerator_Scale < Remainder_Scale
+         Scale_Adjust := Remainder_Scale - Numerator_Scale;
+         Adjusted_Remainder :=
+           Build_Multiply
+             (N, Computed_Remainder, Integer_Literal (N, 10 ** Scale_Adjust));
+      end if;
+
+      --  Assignment of remainder result
+
+      Append_To (Stmts,
+        Make_Assignment_Statement (Loc,
+          Name => Remainder,
+          Expression =>
+            Unchecked_Convert_To (Remainder_Type, Adjusted_Remainder)));
+
+      --  Final step is to rewrite the call with a block containing the
+      --  above sequence of constructed statements for the divide operation.
+
+      Rewrite (N,
+        Make_Block_Statement (Loc,
+          Handled_Statement_Sequence =>
+            Make_Handled_Sequence_Of_Statements (Loc,
+              Statements => Stmts)));
+
+      Analyze (N);
+   end Expand_Decimal_Divide_Call;
+
+   -----------------------------------------------
+   -- Expand_Divide_Fixed_By_Fixed_Giving_Fixed --
+   -----------------------------------------------
+
+   procedure Expand_Divide_Fixed_By_Fixed_Giving_Fixed (N : Node_Id) is
+      Left  : constant Node_Id := Left_Opnd (N);
+      Right : constant Node_Id := Right_Opnd (N);
+
+   begin
+      --  Suppress expansion of a fixed-by-fixed division if the
+      --  operation is supported directly by the target.
+
+      if Target_Has_Fixed_Ops (Etype (Left), Etype (Right), Etype (N)) then
+         return;
+      end if;
+
+      if Etype (Left) = Universal_Real then
+         Do_Divide_Universal_Fixed (N);
+
+      elsif Etype (Right) = Universal_Real then
+         Do_Divide_Fixed_Universal (N);
+
+      else
+         Do_Divide_Fixed_Fixed (N);
+
+         --  A focused optimization: if after constant folding the
+         --  expression is of the form:  T ((Exp * D) / D), where D is
+         --  a static constant, return  T (Exp). This form will show up
+         --  when D is the denominator of the static expression for the
+         --  'small of fixed-point types involved. This transformation
+         --  removes a division that may be expensive on some targets.
+
+         if Nkind (N) = N_Type_Conversion
+           and then Nkind (Expression (N)) = N_Op_Divide
+         then
+            declare
+               Num : constant Node_Id := Left_Opnd  (Expression (N));
+               Den : constant Node_Id := Right_Opnd (Expression (N));
+
+            begin
+               if Nkind (Den) = N_Integer_Literal
+                 and then Nkind (Num) = N_Op_Multiply
+                 and then Nkind (Right_Opnd (Num)) = N_Integer_Literal
+                 and then Intval (Den) = Intval (Right_Opnd (Num))
+               then
+                  Rewrite (Expression (N), Left_Opnd (Num));
+               end if;
+            end;
+         end if;
+      end if;
+   end Expand_Divide_Fixed_By_Fixed_Giving_Fixed;
+
+   -----------------------------------------------
+   -- Expand_Divide_Fixed_By_Fixed_Giving_Float --
+   -----------------------------------------------
+
+   --  The division is done in Universal_Real, and the result is multiplied
+   --  by the small ratio, which is Small (Right) / Small (Left). Special
+   --  treatment is required for universal operands, which represent their
+   --  own value and do not require conversion.
+
+   procedure Expand_Divide_Fixed_By_Fixed_Giving_Float (N : Node_Id) is
+      Left  : constant Node_Id := Left_Opnd (N);
+      Right : constant Node_Id := Right_Opnd (N);
+
+      Left_Type  : constant Entity_Id := Etype (Left);
+      Right_Type : constant Entity_Id := Etype (Right);
+
+   begin
+      --  Case of left operand is universal real, the result we want is:
+
+      --    Left_Value / (Right_Value * Right_Small)
+
+      --  so we compute this as:
+
+      --    (Left_Value / Right_Small) / Right_Value
+
+      if Left_Type = Universal_Real then
+         Set_Result (N,
+           Build_Divide (N,
+             Real_Literal (N, Realval (Left) / Small_Value (Right_Type)),
+             Fpt_Value (Right)));
+
+      --  Case of right operand is universal real, the result we want is
+
+      --    (Left_Value * Left_Small) / Right_Value
+
+      --  so we compute this as:
+
+      --    Left_Value * (Left_Small / Right_Value)
+
+      --  Note we invert to a multiplication since usually floating-point
+      --  multiplication is much faster than floating-point division.
+
+      elsif Right_Type = Universal_Real then
+         Set_Result (N,
+           Build_Multiply (N,
+             Fpt_Value (Left),
+             Real_Literal (N, Small_Value (Left_Type) / Realval (Right))));
+
+      --  Both operands are fixed, so the value we want is
+
+      --    (Left_Value * Left_Small) / (Right_Value * Right_Small)
+
+      --  which we compute as:
+
+      --    (Left_Value / Right_Value) * (Left_Small / Right_Small)
+
+      else
+         Set_Result (N,
+           Build_Multiply (N,
+             Build_Divide (N, Fpt_Value (Left), Fpt_Value (Right)),
+             Real_Literal (N,
+               Small_Value (Left_Type) / Small_Value (Right_Type))));
+      end if;
+   end Expand_Divide_Fixed_By_Fixed_Giving_Float;
+
+   -------------------------------------------------
+   -- Expand_Divide_Fixed_By_Fixed_Giving_Integer --
+   -------------------------------------------------
+
+   procedure Expand_Divide_Fixed_By_Fixed_Giving_Integer (N : Node_Id) is
+      Left  : constant Node_Id := Left_Opnd (N);
+      Right : constant Node_Id := Right_Opnd (N);
+   begin
+      if Etype (Left) = Universal_Real then
+         Do_Divide_Universal_Fixed (N);
+      elsif Etype (Right) = Universal_Real then
+         Do_Divide_Fixed_Universal (N);
+      else
+         Do_Divide_Fixed_Fixed (N);
+      end if;
+   end Expand_Divide_Fixed_By_Fixed_Giving_Integer;
+
+   -------------------------------------------------
+   -- Expand_Divide_Fixed_By_Integer_Giving_Fixed --
+   -------------------------------------------------
+
+   --  Since the operand and result fixed-point type is the same, this is
+   --  a straight divide by the right operand, the small can be ignored.
+
+   procedure Expand_Divide_Fixed_By_Integer_Giving_Fixed (N : Node_Id) is
+      Left  : constant Node_Id := Left_Opnd (N);
+      Right : constant Node_Id := Right_Opnd (N);
+   begin
+      Set_Result (N, Build_Divide (N, Left, Right));
+   end Expand_Divide_Fixed_By_Integer_Giving_Fixed;
+
+   -------------------------------------------------
+   -- Expand_Multiply_Fixed_By_Fixed_Giving_Fixed --
+   -------------------------------------------------
+
+   procedure Expand_Multiply_Fixed_By_Fixed_Giving_Fixed (N : Node_Id) is
+      Left  : constant Node_Id := Left_Opnd (N);
+      Right : constant Node_Id := Right_Opnd (N);
+
+      procedure Rewrite_Non_Static_Universal (Opnd : Node_Id);
+      --  The operand may be a non-static universal value, such an
+      --  exponentiation with a non-static exponent. In that case, treat
+      --  as a fixed * fixed multiplication, and convert the argument to
+      --  the target fixed type.
+
+      ----------------------------------
+      -- Rewrite_Non_Static_Universal --
+      ----------------------------------
+
+      procedure Rewrite_Non_Static_Universal (Opnd : Node_Id) is
+         Loc : constant Source_Ptr := Sloc (N);
+      begin
+         Rewrite (Opnd,
+           Make_Type_Conversion (Loc,
+             Subtype_Mark => New_Occurrence_Of (Etype (N), Loc),
+             Expression   => Expression (Opnd)));
+         Analyze_And_Resolve (Opnd, Etype (N));
+      end Rewrite_Non_Static_Universal;
+
+   --  Start of processing for Expand_Multiply_Fixed_By_Fixed_Giving_Fixed
+
+   begin
+      --  Suppress expansion of a fixed-by-fixed multiplication if the
+      --  operation is supported directly by the target.
+
+      if Target_Has_Fixed_Ops (Etype (Left), Etype (Right), Etype (N)) then
+         return;
+      end if;
+
+      if Etype (Left) = Universal_Real then
+         if Nkind (Left) = N_Real_Literal then
+            Do_Multiply_Fixed_Universal (N, Left => Right, Right => Left);
+
+         elsif Nkind (Left) = N_Type_Conversion then
+            Rewrite_Non_Static_Universal (Left);
+            Do_Multiply_Fixed_Fixed (N);
+         end if;
+
+      elsif Etype (Right) = Universal_Real then
+         if Nkind (Right) = N_Real_Literal then
+            Do_Multiply_Fixed_Universal (N, Left, Right);
+
+         elsif Nkind (Right) = N_Type_Conversion then
+            Rewrite_Non_Static_Universal (Right);
+            Do_Multiply_Fixed_Fixed (N);
+         end if;
+
+      else
+         Do_Multiply_Fixed_Fixed (N);
+      end if;
+   end Expand_Multiply_Fixed_By_Fixed_Giving_Fixed;
+
+   -------------------------------------------------
+   -- Expand_Multiply_Fixed_By_Fixed_Giving_Float --
+   -------------------------------------------------
+
+   --  The multiply is done in Universal_Real, and the result is multiplied
+   --  by the adjustment for the smalls which is Small (Right) * Small (Left).
+   --  Special treatment is required for universal operands.
+
+   procedure Expand_Multiply_Fixed_By_Fixed_Giving_Float (N : Node_Id) is
+      Left  : constant Node_Id := Left_Opnd (N);
+      Right : constant Node_Id := Right_Opnd (N);
+
+      Left_Type  : constant Entity_Id := Etype (Left);
+      Right_Type : constant Entity_Id := Etype (Right);
+
+   begin
+      --  Case of left operand is universal real, the result we want is
+
+      --    Left_Value * (Right_Value * Right_Small)
+
+      --  so we compute this as:
+
+      --    (Left_Value * Right_Small) * Right_Value;
+
+      if Left_Type = Universal_Real then
+         Set_Result (N,
+           Build_Multiply (N,
+             Real_Literal (N, Realval (Left) * Small_Value (Right_Type)),
+             Fpt_Value (Right)));
+
+      --  Case of right operand is universal real, the result we want is
+
+      --    (Left_Value * Left_Small) * Right_Value
+
+      --  so we compute this as:
+
+      --    Left_Value * (Left_Small * Right_Value)
+
+      elsif Right_Type = Universal_Real then
+         Set_Result (N,
+           Build_Multiply (N,
+             Fpt_Value (Left),
+             Real_Literal (N, Small_Value (Left_Type) * Realval (Right))));
+
+      --  Both operands are fixed, so the value we want is
+
+      --    (Left_Value * Left_Small) * (Right_Value * Right_Small)
+
+      --  which we compute as:
+
+      --    (Left_Value * Right_Value) * (Right_Small * Left_Small)
+
+      else
+         Set_Result (N,
+           Build_Multiply (N,
+             Build_Multiply (N, Fpt_Value (Left), Fpt_Value (Right)),
+             Real_Literal (N,
+               Small_Value (Right_Type) * Small_Value (Left_Type))));
+      end if;
+   end Expand_Multiply_Fixed_By_Fixed_Giving_Float;
+
+   ---------------------------------------------------
+   -- Expand_Multiply_Fixed_By_Fixed_Giving_Integer --
+   ---------------------------------------------------
+
+   procedure Expand_Multiply_Fixed_By_Fixed_Giving_Integer (N : Node_Id) is
+      Loc   : constant Source_Ptr := Sloc (N);
+      Left  : constant Node_Id    := Left_Opnd (N);
+      Right : constant Node_Id    := Right_Opnd (N);
+
+   begin
+      if Etype (Left) = Universal_Real then
+         Do_Multiply_Fixed_Universal (N, Left => Right, Right => Left);
+
+      elsif Etype (Right) = Universal_Real then
+         Do_Multiply_Fixed_Universal (N, Left, Right);
+
+      --  If both types are equal and we need to avoid floating point
+      --  instructions, it's worth introducing a temporary with the
+      --  common type, because it may be evaluated more simply without
+      --  the need for run-time use of floating point.
+
+      elsif Etype (Right) = Etype (Left)
+        and then Restriction_Active (No_Floating_Point)
+      then
+         declare
+            Temp : constant Entity_Id := Make_Temporary (Loc, 'F');
+            Mult : constant Node_Id   := Make_Op_Multiply (Loc, Left, Right);
+            Decl : constant Node_Id   :=
+              Make_Object_Declaration (Loc,
+                Defining_Identifier => Temp,
+                Object_Definition   => New_Occurrence_Of (Etype (Right), Loc),
+                Expression          => Mult);
+
+         begin
+            Insert_Action (N, Decl);
+            Rewrite (N,
+              OK_Convert_To (Etype (N), New_Occurrence_Of (Temp, Loc)));
+            Analyze_And_Resolve (N, Standard_Integer);
+         end;
+
+      else
+         Do_Multiply_Fixed_Fixed (N);
+      end if;
+   end Expand_Multiply_Fixed_By_Fixed_Giving_Integer;
+
+   ---------------------------------------------------
+   -- Expand_Multiply_Fixed_By_Integer_Giving_Fixed --
+   ---------------------------------------------------
+
+   --  Since the operand and result fixed-point type is the same, this is
+   --  a straight multiply by the right operand, the small can be ignored.
+
+   procedure Expand_Multiply_Fixed_By_Integer_Giving_Fixed (N : Node_Id) is
+   begin
+      Set_Result (N,
+        Build_Multiply (N, Left_Opnd (N), Right_Opnd (N)));
+   end Expand_Multiply_Fixed_By_Integer_Giving_Fixed;
+
+   ---------------------------------------------------
+   -- Expand_Multiply_Integer_By_Fixed_Giving_Fixed --
+   ---------------------------------------------------
+
+   --  Since the operand and result fixed-point type is the same, this is
+   --  a straight multiply by the right operand, the small can be ignored.
+
+   procedure Expand_Multiply_Integer_By_Fixed_Giving_Fixed (N : Node_Id) is
+   begin
+      Set_Result (N,
+        Build_Multiply (N, Left_Opnd (N), Right_Opnd (N)));
+   end Expand_Multiply_Integer_By_Fixed_Giving_Fixed;
+
+   ---------------
+   -- Fpt_Value --
+   ---------------
+
+   function Fpt_Value (N : Node_Id) return Node_Id is
+      Typ   : constant Entity_Id  := Etype (N);
+
+   begin
+      if Is_Integer_Type (Typ)
+        or else Is_Floating_Point_Type (Typ)
+      then
+         return Build_Conversion (N, Universal_Real, N);
+
+      --  Fixed-point case, must get integer value first
+
+      else
+         return Build_Conversion (N, Universal_Real, N);
+      end if;
+   end Fpt_Value;
+
+   ---------------------
+   -- Integer_Literal --
+   ---------------------
+
+   function Integer_Literal
+     (N        : Node_Id;
+      V        : Uint;
+      Negative : Boolean := False) return Node_Id
+   is
+      T : Entity_Id;
+      L : Node_Id;
+
+   begin
+      if V < Uint_2 ** 7 then
+         T := Standard_Integer_8;
+
+      elsif V < Uint_2 ** 15 then
+         T := Standard_Integer_16;
+
+      elsif V < Uint_2 ** 31 then
+         T := Standard_Integer_32;
+
+      elsif V < Uint_2 ** 63 then
+         T := Standard_Integer_64;
+
+      else
+         return Empty;
+      end if;
+
+      if Negative then
+         L := Make_Integer_Literal (Sloc (N), UI_Negate (V));
+      else
+         L := Make_Integer_Literal (Sloc (N), V);
+      end if;
+
+      --  Set type of result in case used elsewhere (see note at start)
+
+      Set_Etype (L, T);
+      Set_Is_Static_Expression (L);
+
+      --  We really need to set Analyzed here because we may be creating a
+      --  very strange beast, namely an integer literal typed as fixed-point
+      --  and the analyzer won't like that. Probably we should allow the
+      --  Treat_Fixed_As_Integer flag to appear on integer literal nodes
+      --  and teach the analyzer how to handle them ???
+
+      Set_Analyzed (L);
+      return L;
+   end Integer_Literal;
+
+   ------------------
+   -- Real_Literal --
+   ------------------
+
+   function Real_Literal (N : Node_Id; V : Ureal) return Node_Id is
+      L : Node_Id;
+
+   begin
+      L := Make_Real_Literal (Sloc (N), V);
+
+      --  Set type of result in case used elsewhere (see note at start)
+
+      Set_Etype (L, Universal_Real);
+      return L;
+   end Real_Literal;
+
+   ------------------------
+   -- Rounded_Result_Set --
+   ------------------------
+
+   function Rounded_Result_Set (N : Node_Id) return Boolean is
+      K : constant Node_Kind := Nkind (N);
+   begin
+      if (K = N_Type_Conversion or else
+          K = N_Op_Divide       or else
+          K = N_Op_Multiply)
+        and then
+          (Rounded_Result (N) or else Is_Integer_Type (Etype (N)))
+      then
+         return True;
+      else
+         return False;
+      end if;
+   end Rounded_Result_Set;
+
+   ----------------
+   -- Set_Result --
+   ----------------
+
+   procedure Set_Result
+     (N     : Node_Id;
+      Expr  : Node_Id;
+      Rchk  : Boolean := False;
+      Trunc : Boolean := False)
+   is
+      Cnode : Node_Id;
+
+      Expr_Type   : constant Entity_Id := Etype (Expr);
+      Result_Type : constant Entity_Id := Etype (N);
+
+   begin
+      --  No conversion required if types match and no range check or truncate
+
+      if Result_Type = Expr_Type and then not (Rchk or Trunc) then
+         Cnode := Expr;
+
+      --  Else perform required conversion
+
+      else
+         Cnode := Build_Conversion (N, Result_Type, Expr, Rchk, Trunc);
+      end if;
+
+      Rewrite (N, Cnode);
+      Analyze_And_Resolve (N, Result_Type);
+   end Set_Result;
+
+end Exp_Fixd;