1------------------------------------------------------------------------------
2--                                                                          --
3--                         GNAT COMPILER COMPONENTS                         --
4--                                                                          --
5--                       S Y S T E M . B I G N U M S                        --
6--                                                                          --
7--                                 B o d y                                  --
8--                                                                          --
9--            Copyright (C) 2012, Free Software Foundation, Inc.            --
10--                                                                          --
11-- GNAT is free software;  you can  redistribute it  and/or modify it under --
12-- terms of the  GNU General Public License as published  by the Free Soft- --
13-- ware  Foundation;  either version 3,  or (at your option) any later ver- --
14-- sion.  GNAT is distributed in the hope that it will be useful, but WITH- --
15-- OUT ANY WARRANTY;  without even the  implied warranty of MERCHANTABILITY --
16-- or FITNESS FOR A PARTICULAR PURPOSE.                                     --
17--                                                                          --
18-- As a special exception under Section 7 of GPL version 3, you are granted --
19-- additional permissions described in the GCC Runtime Library Exception,   --
20-- version 3.1, as published by the Free Software Foundation.               --
21--                                                                          --
22-- You should have received a copy of the GNU General Public License and    --
23-- a copy of the GCC Runtime Library Exception along with this program;     --
24-- see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see    --
25-- <http://www.gnu.org/licenses/>.                                          --
26--                                                                          --
27-- GNAT was originally developed  by the GNAT team at  New York University. --
28-- Extensive contributions were provided by Ada Core Technologies Inc.      --
29--                                                                          --
30------------------------------------------------------------------------------
31
32--  This package provides arbitrary precision signed integer arithmetic for
33--  use in computing intermediate values in expressions for the case where
34--  pragma Overflow_Check (Eliminate) is in effect.
35
36with System;                  use System;
37with System.Secondary_Stack;  use System.Secondary_Stack;
38with System.Storage_Elements; use System.Storage_Elements;
39
40package body System.Bignums is
41
42   use Interfaces;
43   --  So that operations on Unsigned_32 are available
44
45   type DD is mod Base ** 2;
46   --  Double length digit used for intermediate computations
47
48   function MSD (X : DD) return SD is (SD (X / Base));
49   function LSD (X : DD) return SD is (SD (X mod Base));
50   --  Most significant and least significant digit of double digit value
51
52   function "&" (X, Y : SD) return DD is (DD (X) * Base + DD (Y));
53   --  Compose double digit value from two single digit values
54
55   subtype LLI is Long_Long_Integer;
56
57   One_Data : constant Digit_Vector (1 .. 1) := (1 => 1);
58   --  Constant one
59
60   Zero_Data : constant Digit_Vector (1 .. 0) := (1 .. 0 => 0);
61   --  Constant zero
62
63   -----------------------
64   -- Local Subprograms --
65   -----------------------
66
67   function Add (X, Y : Digit_Vector; X_Neg, Y_Neg : Boolean) return Bignum
68   with Pre => X'First = 1 and then Y'First = 1;
69   --  This procedure adds two signed numbers returning the Sum, it is used
70   --  for both addition and subtraction. The value computed is X + Y, with
71   --  X_Neg and Y_Neg giving the signs of the operands.
72
73   function Allocate_Bignum (Len : Length) return Bignum
74   with Post => Allocate_Bignum'Result.Len = Len;
75   --  Allocate Bignum value of indicated length on secondary stack. On return
76   --  the Neg and D fields are left uninitialized.
77
78   type Compare_Result is (LT, EQ, GT);
79   --  Indicates result of comparison in following call
80
81   function Compare
82     (X, Y         : Digit_Vector;
83      X_Neg, Y_Neg : Boolean) return Compare_Result
84   with Pre => X'First = 1 and then Y'First = 1;
85   --  Compare (X with sign X_Neg) with (Y with sign Y_Neg), and return the
86   --  result of the signed comparison.
87
88   procedure Div_Rem
89     (X, Y              : Bignum;
90      Quotient          : out Bignum;
91      Remainder         : out Bignum;
92      Discard_Quotient  : Boolean := False;
93      Discard_Remainder : Boolean := False);
94   --  Returns the Quotient and Remainder from dividing abs (X) by abs (Y). The
95   --  values of X and Y are not modified. If Discard_Quotient is True, then
96   --  Quotient is undefined on return, and if Discard_Remainder is True, then
97   --  Remainder is undefined on return. Service routine for Big_Div/Rem/Mod.
98
99   procedure Free_Bignum (X : Bignum) is null;
100   --  Called to free a Bignum value used in intermediate computations. In
101   --  this implementation using the secondary stack, it does nothing at all,
102   --  because we rely on Mark/Release, but it may be of use for some
103   --  alternative implementation.
104
105   function Normalize
106     (X   : Digit_Vector;
107      Neg : Boolean := False) return Bignum;
108   --  Given a digit vector and sign, allocate and construct a Bignum value.
109   --  Note that X may have leading zeroes which must be removed, and if the
110   --  result is zero, the sign is forced positive.
111
112   ---------
113   -- Add --
114   ---------
115
116   function Add (X, Y : Digit_Vector; X_Neg, Y_Neg : Boolean) return Bignum is
117   begin
118      --  If signs are the same, we are doing an addition, it is convenient to
119      --  ensure that the first operand is the longer of the two.
120
121      if X_Neg = Y_Neg then
122         if X'Last < Y'Last then
123            return Add (X => Y, Y => X, X_Neg => Y_Neg, Y_Neg => X_Neg);
124
125         --  Here signs are the same, and the first operand is the longer
126
127         else
128            pragma Assert (X_Neg = Y_Neg and then X'Last >= Y'Last);
129
130            --  Do addition, putting result in Sum (allowing for carry)
131
132            declare
133               Sum : Digit_Vector (0 .. X'Last);
134               RD  : DD;
135
136            begin
137               RD := 0;
138               for J in reverse 1 .. X'Last loop
139                  RD := RD + DD (X (J));
140
141                  if J >= 1 + (X'Last - Y'Last)  then
142                     RD := RD + DD (Y (J - (X'Last - Y'Last)));
143                  end if;
144
145                  Sum (J) := LSD (RD);
146                  RD := RD / Base;
147               end loop;
148
149               Sum (0) := SD (RD);
150               return Normalize (Sum, X_Neg);
151            end;
152         end if;
153
154      --  Signs are different so really this is a subtraction, we want to make
155      --  sure that the largest magnitude operand is the first one, and then
156      --  the result will have the sign of the first operand.
157
158      else
159         declare
160            CR : constant Compare_Result := Compare (X, Y, False, False);
161
162         begin
163            if CR = EQ then
164               return Normalize (Zero_Data);
165
166            elsif CR = LT then
167               return Add (X => Y, Y => X, X_Neg => Y_Neg, Y_Neg => X_Neg);
168
169            else
170               pragma Assert (X_Neg /= Y_Neg and then CR = GT);
171
172               --  Do subtraction, putting result in Diff
173
174               declare
175                  Diff : Digit_Vector (1 .. X'Length);
176                  RD   : DD;
177
178               begin
179                  RD := 0;
180                  for J in reverse 1 .. X'Last loop
181                     RD := RD + DD (X (J));
182
183                     if J >= 1 + (X'Last - Y'Last)  then
184                        RD := RD - DD (Y (J - (X'Last - Y'Last)));
185                     end if;
186
187                     Diff (J) := LSD (RD);
188                     RD := (if RD < Base then 0 else -1);
189                  end loop;
190
191                  return Normalize (Diff, X_Neg);
192               end;
193            end if;
194         end;
195      end if;
196   end Add;
197
198   ---------------------
199   -- Allocate_Bignum --
200   ---------------------
201
202   function Allocate_Bignum (Len : Length) return Bignum is
203      Addr : Address;
204
205   begin
206      --  Change the if False here to if True to get allocation on the heap
207      --  instead of the secondary stack, which is convenient for debugging
208      --  System.Bignum itself.
209
210      if False then
211         declare
212            B : Bignum;
213         begin
214            B := new Bignum_Data'(Len, False, (others => 0));
215            return B;
216         end;
217
218      --  Normal case of allocation on the secondary stack
219
220      else
221         --  Note: The approach used here is designed to avoid strict aliasing
222         --  warnings that appeared previously using unchecked conversion.
223
224         SS_Allocate (Addr, Storage_Offset (4 + 4 * Len));
225
226         declare
227            B : Bignum;
228            for B'Address use Addr'Address;
229            pragma Import (Ada, B);
230
231            BD : Bignum_Data (Len);
232            for BD'Address use Addr;
233            pragma Import (Ada, BD);
234
235            --  Expose a writable view of discriminant BD.Len so that we can
236            --  initialize it. We need to use the exact layout of the record
237            --  to ensure that the Length field has 24 bits as expected.
238
239            type Bignum_Data_Header is record
240               Len : Length;
241               Neg : Boolean;
242            end record;
243
244            for Bignum_Data_Header use record
245               Len at 0 range 0 .. 23;
246               Neg at 3 range 0 .. 7;
247            end record;
248
249            BDH : Bignum_Data_Header;
250            for BDH'Address use BD'Address;
251            pragma Import (Ada, BDH);
252
253            pragma Assert (BDH.Len'Size = BD.Len'Size);
254
255         begin
256            BDH.Len := Len;
257            return B;
258         end;
259      end if;
260   end Allocate_Bignum;
261
262   -------------
263   -- Big_Abs --
264   -------------
265
266   function Big_Abs (X : Bignum) return Bignum is
267   begin
268      return Normalize (X.D);
269   end Big_Abs;
270
271   -------------
272   -- Big_Add --
273   -------------
274
275   function Big_Add  (X, Y : Bignum) return Bignum is
276   begin
277      return Add (X.D, Y.D, X.Neg, Y.Neg);
278   end Big_Add;
279
280   -------------
281   -- Big_Div --
282   -------------
283
284   --  This table is excerpted from RM 4.5.5(28-30) and shows how the result
285   --  varies with the signs of the operands.
286
287   --   A      B   A/B      A     B    A/B
288   --
289   --   10     5    2      -10    5    -2
290   --   11     5    2      -11    5    -2
291   --   12     5    2      -12    5    -2
292   --   13     5    2      -13    5    -2
293   --   14     5    2      -14    5    -2
294   --
295   --   A      B   A/B      A     B    A/B
296   --
297   --   10    -5   -2      -10   -5     2
298   --   11    -5   -2      -11   -5     2
299   --   12    -5   -2      -12   -5     2
300   --   13    -5   -2      -13   -5     2
301   --   14    -5   -2      -14   -5     2
302
303   function Big_Div  (X, Y : Bignum) return Bignum is
304      Q, R : Bignum;
305   begin
306      Div_Rem (X, Y, Q, R, Discard_Remainder => True);
307      Q.Neg := Q.Len > 0 and then (X.Neg xor Y.Neg);
308      return Q;
309   end Big_Div;
310
311   -------------
312   -- Big_Exp --
313   -------------
314
315   function Big_Exp  (X, Y : Bignum) return Bignum is
316
317      function "**" (X : Bignum; Y : SD) return Bignum;
318      --  Internal routine where we know right operand is one word
319
320      ----------
321      -- "**" --
322      ----------
323
324      function "**" (X : Bignum; Y : SD) return Bignum is
325      begin
326         case Y is
327
328            --  X ** 0 is 1
329
330            when 0 =>
331               return Normalize (One_Data);
332
333            --  X ** 1 is X
334
335            when 1 =>
336               return Normalize (X.D);
337
338            --  X ** 2 is X * X
339
340            when 2 =>
341               return Big_Mul (X, X);
342
343            --  For X greater than 2, use the recursion
344
345            --  X even, X ** Y = (X ** (Y/2)) ** 2;
346            --  X odd,  X ** Y = (X ** (Y/2)) ** 2 * X;
347
348            when others =>
349               declare
350                  XY2  : constant Bignum := X ** (Y / 2);
351                  XY2S : constant Bignum := Big_Mul (XY2, XY2);
352                  Res  : Bignum;
353
354               begin
355                  Free_Bignum (XY2);
356
357                  --  Raise storage error if intermediate value is getting too
358                  --  large, which we arbitrarily define as 200 words for now!
359
360                  if XY2S.Len > 200 then
361                     Free_Bignum (XY2S);
362                     raise Storage_Error with
363                       "exponentiation result is too large";
364                  end if;
365
366                  --  Otherwise take care of even/odd cases
367
368                  if (Y and 1) = 0 then
369                     return XY2S;
370
371                  else
372                     Res := Big_Mul (XY2S, X);
373                     Free_Bignum (XY2S);
374                     return Res;
375                  end if;
376               end;
377         end case;
378      end "**";
379
380   --  Start of processing for Big_Exp
381
382   begin
383      --  Error if right operand negative
384
385      if Y.Neg then
386         raise Constraint_Error with "exponentiation to negative power";
387
388      --  X ** 0 is always 1 (including 0 ** 0, so do this test first)
389
390      elsif Y.Len = 0 then
391         return Normalize (One_Data);
392
393      --  0 ** X is always 0 (for X non-zero)
394
395      elsif X.Len = 0 then
396         return Normalize (Zero_Data);
397
398      --  (+1) ** Y = 1
399      --  (-1) ** Y = +/-1 depending on whether Y is even or odd
400
401      elsif X.Len = 1 and then X.D (1) = 1 then
402         return Normalize
403           (X.D, Neg => X.Neg and then ((Y.D (Y.Len) and 1) = 1));
404
405      --  If the absolute value of the base is greater than 1, then the
406      --  exponent must not be bigger than one word, otherwise the result
407      --  is ludicrously large, and we just signal Storage_Error right away.
408
409      elsif Y.Len > 1 then
410         raise Storage_Error with "exponentiation result is too large";
411
412      --  Special case (+/-)2 ** K, where K is 1 .. 31 using a shift
413
414      elsif X.Len = 1 and then X.D (1) = 2 and then Y.D (1) < 32 then
415         declare
416            D : constant Digit_Vector (1 .. 1) :=
417                  (1 => Shift_Left (SD'(1), Natural (Y.D (1))));
418         begin
419            return Normalize (D, X.Neg);
420         end;
421
422      --  Remaining cases have right operand of one word
423
424      else
425         return X ** Y.D (1);
426      end if;
427   end Big_Exp;
428
429   ------------
430   -- Big_EQ --
431   ------------
432
433   function Big_EQ (X, Y : Bignum) return Boolean is
434   begin
435      return Compare (X.D, Y.D, X.Neg, Y.Neg) = EQ;
436   end Big_EQ;
437
438   ------------
439   -- Big_GE --
440   ------------
441
442   function Big_GE (X, Y : Bignum) return Boolean is
443   begin
444      return Compare (X.D, Y.D, X.Neg, Y.Neg) /= LT;
445   end Big_GE;
446
447   ------------
448   -- Big_GT --
449   ------------
450
451   function Big_GT (X, Y : Bignum) return Boolean is
452   begin
453      return Compare (X.D, Y.D, X.Neg, Y.Neg) = GT;
454   end Big_GT;
455
456   ------------
457   -- Big_LE --
458   ------------
459
460   function Big_LE (X, Y : Bignum) return Boolean is
461   begin
462      return Compare (X.D, Y.D, X.Neg, Y.Neg) /= GT;
463   end Big_LE;
464
465   ------------
466   -- Big_LT --
467   ------------
468
469   function Big_LT (X, Y : Bignum) return Boolean is
470   begin
471      return Compare (X.D, Y.D, X.Neg, Y.Neg) = LT;
472   end Big_LT;
473
474   -------------
475   -- Big_Mod --
476   -------------
477
478   --  This table is excerpted from RM 4.5.5(28-30) and shows how the result
479   --  of Rem and Mod vary with the signs of the operands.
480
481   --   A      B    A mod B  A rem B     A     B    A mod B  A rem B
482
483   --   10     5       0        0       -10    5       0        0
484   --   11     5       1        1       -11    5       4       -1
485   --   12     5       2        2       -12    5       3       -2
486   --   13     5       3        3       -13    5       2       -3
487   --   14     5       4        4       -14    5       1       -4
488
489   --   A      B    A mod B  A rem B     A     B    A mod B  A rem B
490
491   --   10    -5       0        0       -10   -5       0        0
492   --   11    -5      -4        1       -11   -5      -1       -1
493   --   12    -5      -3        2       -12   -5      -2       -2
494   --   13    -5      -2        3       -13   -5      -3       -3
495   --   14    -5      -1        4       -14   -5      -4       -4
496
497   function Big_Mod (X, Y : Bignum) return Bignum is
498      Q, R : Bignum;
499
500   begin
501      --  If signs are same, result is same as Rem
502
503      if X.Neg = Y.Neg then
504         return Big_Rem (X, Y);
505
506      --  Case where Mod is different
507
508      else
509         --  Do division
510
511         Div_Rem (X, Y, Q, R, Discard_Quotient => True);
512
513         --  Zero result is unchanged
514
515         if R.Len = 0 then
516            return R;
517
518         --  Otherwise adjust result
519
520         else
521            declare
522               T1 : constant Bignum := Big_Sub (Y, R);
523            begin
524               T1.Neg := Y.Neg;
525               Free_Bignum (R);
526               return T1;
527            end;
528         end if;
529      end if;
530   end Big_Mod;
531
532   -------------
533   -- Big_Mul --
534   -------------
535
536   function Big_Mul (X, Y : Bignum) return Bignum is
537      Result : Digit_Vector (1 .. X.Len + Y.Len) := (others => 0);
538      --  Accumulate result (max length of result is sum of operand lengths)
539
540      L : Length;
541      --  Current result digit
542
543      D : DD;
544      --  Result digit
545
546   begin
547      for J in 1 .. X.Len loop
548         for K in 1 .. Y.Len loop
549            L := Result'Last - (X.Len - J) - (Y.Len - K);
550            D := DD (X.D (J)) * DD (Y.D (K)) + DD (Result (L));
551            Result (L) := LSD (D);
552            D := D / Base;
553
554            --  D is carry which must be propagated
555
556            while D /= 0 and then L >= 1 loop
557               L := L - 1;
558               D := D + DD (Result (L));
559               Result (L) := LSD (D);
560               D := D / Base;
561            end loop;
562
563            --  Must not have a carry trying to extend max length
564
565            pragma Assert (D = 0);
566         end loop;
567      end loop;
568
569      --  Return result
570
571      return Normalize (Result, X.Neg xor Y.Neg);
572   end Big_Mul;
573
574   ------------
575   -- Big_NE --
576   ------------
577
578   function Big_NE (X, Y : Bignum) return Boolean is
579   begin
580      return Compare (X.D, Y.D, X.Neg, Y.Neg) /= EQ;
581   end Big_NE;
582
583   -------------
584   -- Big_Neg --
585   -------------
586
587   function Big_Neg (X : Bignum) return Bignum is
588   begin
589      return Normalize (X.D, not X.Neg);
590   end Big_Neg;
591
592   -------------
593   -- Big_Rem --
594   -------------
595
596   --  This table is excerpted from RM 4.5.5(28-30) and shows how the result
597   --  varies with the signs of the operands.
598
599   --   A      B   A rem B   A     B   A rem B
600
601   --   10     5      0     -10    5      0
602   --   11     5      1     -11    5     -1
603   --   12     5      2     -12    5     -2
604   --   13     5      3     -13    5     -3
605   --   14     5      4     -14    5     -4
606
607   --   A      B  A rem B    A     B   A rem B
608
609   --   10    -5     0      -10   -5      0
610   --   11    -5     1      -11   -5     -1
611   --   12    -5     2      -12   -5     -2
612   --   13    -5     3      -13   -5     -3
613   --   14    -5     4      -14   -5     -4
614
615   function Big_Rem (X, Y : Bignum) return Bignum is
616      Q, R : Bignum;
617   begin
618      Div_Rem (X, Y, Q, R, Discard_Quotient => True);
619      R.Neg := R.Len > 0 and then X.Neg;
620      return R;
621   end Big_Rem;
622
623   -------------
624   -- Big_Sub --
625   -------------
626
627   function Big_Sub (X, Y : Bignum) return Bignum is
628   begin
629      --  If right operand zero, return left operand (avoiding sharing)
630
631      if Y.Len = 0 then
632         return Normalize (X.D, X.Neg);
633
634      --  Otherwise add negative of right operand
635
636      else
637         return Add (X.D, Y.D, X.Neg, not Y.Neg);
638      end if;
639   end Big_Sub;
640
641   -------------
642   -- Compare --
643   -------------
644
645   function Compare
646     (X, Y         : Digit_Vector;
647      X_Neg, Y_Neg : Boolean) return Compare_Result
648   is
649   begin
650      --  Signs are different, that's decisive, since 0 is always plus
651
652      if X_Neg /= Y_Neg then
653         return (if X_Neg then LT else GT);
654
655      --  Lengths are different, that's decisive since no leading zeroes
656
657      elsif X'Last /= Y'Last then
658         return (if (X'Last > Y'Last) xor X_Neg then GT else LT);
659
660      --  Need to compare data
661
662      else
663         for J in X'Range loop
664            if X (J) /= Y (J) then
665               return (if (X (J) > Y (J)) xor X_Neg then GT else LT);
666            end if;
667         end loop;
668
669         return EQ;
670      end if;
671   end Compare;
672
673   -------------
674   -- Div_Rem --
675   -------------
676
677   procedure Div_Rem
678     (X, Y              : Bignum;
679      Quotient          : out Bignum;
680      Remainder         : out Bignum;
681      Discard_Quotient  : Boolean := False;
682      Discard_Remainder : Boolean := False)
683   is
684   begin
685      --  Error if division by zero
686
687      if Y.Len = 0 then
688         raise Constraint_Error with "division by zero";
689      end if;
690
691      --  Handle simple cases with special tests
692
693      --  If X < Y then quotient is zero and remainder is X
694
695      if Compare (X.D, Y.D, False, False) = LT then
696         Remainder := Normalize (X.D);
697         Quotient  := Normalize (Zero_Data);
698         return;
699
700      --  If both X and Y are less than 2**63-1, we can use Long_Long_Integer
701      --  arithmetic. Note it is good not to do an accurate range check against
702      --  Long_Long_Integer since -2**63 / -1 overflows!
703
704      elsif (X.Len <= 1 or else (X.Len = 2 and then X.D (1) < 2**31))
705              and then
706            (Y.Len <= 1 or else (Y.Len = 2 and then Y.D (1) < 2**31))
707      then
708         declare
709            A : constant LLI := abs (From_Bignum (X));
710            B : constant LLI := abs (From_Bignum (Y));
711         begin
712            Quotient  := To_Bignum (A / B);
713            Remainder := To_Bignum (A rem B);
714            return;
715         end;
716
717      --  Easy case if divisor is one digit
718
719      elsif Y.Len = 1 then
720         declare
721            ND  : DD;
722            Div : constant DD := DD (Y.D (1));
723
724            Result : Digit_Vector (1 .. X.Len);
725            Remdr  : Digit_Vector (1 .. 1);
726
727         begin
728            ND := 0;
729            for J in 1 .. X.Len loop
730               ND := Base * ND + DD (X.D (J));
731               Result (J) := SD (ND / Div);
732               ND := ND rem Div;
733            end loop;
734
735            Quotient  := Normalize (Result);
736            Remdr (1) := SD (ND);
737            Remainder := Normalize (Remdr);
738            return;
739         end;
740      end if;
741
742      --  The complex full multi-precision case. We will employ algorithm
743      --  D defined in the section "The Classical Algorithms" (sec. 4.3.1)
744      --  of Donald Knuth's "The Art of Computer Programming", Vol. 2, 2nd
745      --  edition. The terminology is adjusted for this section to match that
746      --  reference.
747
748      --  We are dividing X.Len digits of X (called u here) by Y.Len digits
749      --  of Y (called v here), developing the quotient and remainder. The
750      --  numbers are represented using Base, which was chosen so that we have
751      --  the operations of multiplying to single digits (SD) to form a double
752      --  digit (DD), and dividing a double digit (DD) by a single digit (SD)
753      --  to give a single digit quotient and a single digit remainder.
754
755      --  Algorithm D from Knuth
756
757      --  Comments here with square brackets are directly from Knuth
758
759      Algorithm_D : declare
760
761         --  The following lower case variables correspond exactly to the
762         --  terminology used in algorithm D.
763
764         m : constant Length := X.Len - Y.Len;
765         n : constant Length := Y.Len;
766         b : constant DD     := Base;
767
768         u : Digit_Vector (0 .. m + n);
769         v : Digit_Vector (1 .. n);
770         q : Digit_Vector (0 .. m);
771         r : Digit_Vector (1 .. n);
772
773         u0 : SD renames u (0);
774         v1 : SD renames v (1);
775         v2 : SD renames v (2);
776
777         d    : DD;
778         j    : Length;
779         qhat : DD;
780         rhat : DD;
781         temp : DD;
782
783      begin
784         --  Initialize data of left and right operands
785
786         for J in 1 .. m + n loop
787            u (J) := X.D (J);
788         end loop;
789
790         for J in 1 .. n loop
791            v (J) := Y.D (J);
792         end loop;
793
794         --  [Division of nonnegative integers.] Given nonnegative integers u
795         --  = (ul,u2..um+n) and v = (v1,v2..vn), where v1 /= 0 and n > 1, we
796         --  form the quotient u / v = (q0,ql..qm) and the remainder u mod v =
797         --  (r1,r2..rn).
798
799         pragma Assert (v1 /= 0);
800         pragma Assert (n > 1);
801
802         --  Dl. [Normalize.] Set d = b/(vl + 1). Then set (u0,u1,u2..um+n)
803         --  equal to (u1,u2..um+n) times d, and set (v1,v2..vn) equal to
804         --  (v1,v2..vn) times d. Note the introduction of a new digit position
805         --  u0 at the left of u1; if d = 1 all we need to do in this step is
806         --  to set u0 = 0.
807
808         d := b / (DD (v1) + 1);
809
810         if d = 1 then
811            u0 := 0;
812
813         else
814            declare
815               Carry : DD;
816               Tmp   : DD;
817
818            begin
819               --  Multiply Dividend (u) by d
820
821               Carry := 0;
822               for J in reverse 1 .. m + n loop
823                  Tmp   := DD (u (J)) * d + Carry;
824                  u (J) := LSD (Tmp);
825                  Carry := Tmp / Base;
826               end loop;
827
828               u0 := SD (Carry);
829
830               --  Multiply Divisor (v) by d
831
832               Carry := 0;
833               for J in reverse 1 .. n loop
834                  Tmp    := DD (v (J)) * d + Carry;
835                  v (J)  := LSD (Tmp);
836                  Carry  := Tmp / Base;
837               end loop;
838
839               pragma Assert (Carry = 0);
840            end;
841         end if;
842
843         --  D2. [Initialize j.] Set j = 0. The loop on j, steps D2 through D7,
844         --  will be essentially a division of (uj, uj+1..uj+n) by (v1,v2..vn)
845         --  to get a single quotient digit qj.
846
847         j := 0;
848
849         --  Loop through digits
850
851         loop
852            --  Note: In the original printing, step D3 was as follows:
853
854            --  D3. [Calculate qhat.] If uj = v1, set qhat to b-l; otherwise
855            --  set qhat to (uj,uj+1)/v1. Now test if v2 * qhat is greater than
856            --  (uj*b + uj+1 - qhat*v1)*b + uj+2. If so, decrease qhat by 1 and
857            --  repeat this test
858
859            --  This had a bug not discovered till 1995, see Vol 2 errata:
860            --  http://www-cs-faculty.stanford.edu/~uno/err2-2e.ps.gz. Under
861            --  rare circumstances the expression in the test could overflow.
862            --  This version was further corrected in 2005, see Vol 2 errata:
863            --  http://www-cs-faculty.stanford.edu/~uno/all2-pre.ps.gz.
864            --  The code below is the fixed version of this step.
865
866            --  D3. [Calculate qhat.] Set qhat to (uj,uj+1)/v1 and rhat to
867            --  to (uj,uj+1) mod v1.
868
869            temp := u (j) & u (j + 1);
870            qhat := temp / DD (v1);
871            rhat := temp mod DD (v1);
872
873            --  D3 (continued). Now test if qhat >= b or v2*qhat > (rhat,uj+2):
874            --  if so, decrease qhat by 1, increase rhat by v1, and repeat this
875            --  test if rhat < b. [The test on v2 determines at at high speed
876            --  most of the cases in which the trial value qhat is one too
877            --  large, and eliminates all cases where qhat is two too large.]
878
879            while qhat >= b
880              or else DD (v2) * qhat > LSD (rhat) & u (j + 2)
881            loop
882               qhat := qhat - 1;
883               rhat := rhat + DD (v1);
884               exit when rhat >= b;
885            end loop;
886
887            --  D4. [Multiply and subtract.] Replace (uj,uj+1..uj+n) by
888            --  (uj,uj+1..uj+n) minus qhat times (v1,v2..vn). This step
889            --  consists of a simple multiplication by a one-place number,
890            --  combined with a subtraction.
891
892            --  The digits (uj,uj+1..uj+n) are always kept positive; if the
893            --  result of this step is actually negative then (uj,uj+1..uj+n)
894            --  is left as the true value plus b**(n+1), i.e. as the b's
895            --  complement of the true value, and a "borrow" to the left is
896            --  remembered.
897
898            declare
899               Borrow : SD;
900               Carry  : DD;
901               Temp   : DD;
902
903               Negative : Boolean;
904               --  Records if subtraction causes a negative result, requiring
905               --  an add back (case where qhat turned out to be 1 too large).
906
907            begin
908               Borrow := 0;
909               for K in reverse 1 .. n loop
910                  Temp := qhat * DD (v (K)) + DD (Borrow);
911                  Borrow := MSD (Temp);
912
913                  if LSD (Temp) > u (j + K) then
914                     Borrow := Borrow + 1;
915                  end if;
916
917                  u (j + K) := u (j + K) - LSD (Temp);
918               end loop;
919
920               Negative := u (j) < Borrow;
921               u (j) := u (j) - Borrow;
922
923               --  D5. [Test remainder.] Set qj = qhat. If the result of step
924               --  D4 was negative, we will do the add back step (step D6).
925
926               q (j) := LSD (qhat);
927
928               if Negative then
929
930                  --  D6. [Add back.] Decrease qj by 1, and add (0,v1,v2..vn)
931                  --  to (uj,uj+1,uj+2..uj+n). (A carry will occur to the left
932                  --  of uj, and it is be ignored since it cancels with the
933                  --  borrow that occurred in D4.)
934
935                  q (j) := q (j) - 1;
936
937                  Carry := 0;
938                  for K in reverse 1 .. n loop
939                     Temp := DD (v (K)) + DD (u (j + K)) + Carry;
940                     u (j + K) := LSD (Temp);
941                     Carry := Temp / Base;
942                  end loop;
943
944                  u (j) := u (j) + SD (Carry);
945               end if;
946            end;
947
948            --  D7. [Loop on j.] Increase j by one. Now if j <= m, go back to
949            --  D3 (the start of the loop on j).
950
951            j := j + 1;
952            exit when not (j <= m);
953         end loop;
954
955         --  D8. [Unnormalize.] Now (qo,ql..qm) is the desired quotient, and
956         --  the desired remainder may be obtained by dividing (um+1..um+n)
957         --  by d.
958
959         if not Discard_Quotient then
960            Quotient := Normalize (q);
961         end if;
962
963         if not Discard_Remainder then
964            declare
965               Remdr : DD;
966
967            begin
968               Remdr := 0;
969               for K in 1 .. n loop
970                  Remdr := Base * Remdr + DD (u (m + K));
971                  r (K) := SD (Remdr / d);
972                  Remdr := Remdr rem d;
973               end loop;
974
975               pragma Assert (Remdr = 0);
976            end;
977
978            Remainder := Normalize (r);
979         end if;
980      end Algorithm_D;
981   end Div_Rem;
982
983   -----------------
984   -- From_Bignum --
985   -----------------
986
987   function From_Bignum (X : Bignum) return Long_Long_Integer is
988   begin
989      if X.Len = 0 then
990         return 0;
991
992      elsif X.Len = 1 then
993         return (if X.Neg then -LLI (X.D (1)) else LLI (X.D (1)));
994
995      elsif X.Len = 2 then
996         declare
997            Mag : constant DD := X.D (1) & X.D (2);
998         begin
999            if X.Neg and then Mag <= 2 ** 63 then
1000               return -LLI (Mag);
1001            elsif Mag < 2 ** 63 then
1002               return LLI (Mag);
1003            end if;
1004         end;
1005      end if;
1006
1007      raise Constraint_Error with "expression value out of range";
1008   end From_Bignum;
1009
1010   -------------------------
1011   -- Bignum_In_LLI_Range --
1012   -------------------------
1013
1014   function Bignum_In_LLI_Range (X : Bignum) return Boolean is
1015   begin
1016      --  If length is 0 or 1, definitely fits
1017
1018      if X.Len <= 1 then
1019         return True;
1020
1021      --  If length is greater than 2, definitely does not fit
1022
1023      elsif X.Len > 2 then
1024         return False;
1025
1026      --  Length is 2, more tests needed
1027
1028      else
1029         declare
1030            Mag : constant DD := X.D (1) & X.D (2);
1031         begin
1032            return Mag < 2 ** 63 or else (X.Neg and then Mag = 2 ** 63);
1033         end;
1034      end if;
1035   end Bignum_In_LLI_Range;
1036
1037   ---------------
1038   -- Normalize --
1039   ---------------
1040
1041   function Normalize
1042     (X   : Digit_Vector;
1043      Neg : Boolean := False) return Bignum
1044   is
1045      B : Bignum;
1046      J : Length;
1047
1048   begin
1049      J := X'First;
1050      while J <= X'Last and then X (J) = 0 loop
1051         J := J + 1;
1052      end loop;
1053
1054      B := Allocate_Bignum (X'Last - J + 1);
1055      B.Neg := B.Len > 0 and then Neg;
1056      B.D := X (J .. X'Last);
1057      return B;
1058   end Normalize;
1059
1060   ---------------
1061   -- To_Bignum --
1062   ---------------
1063
1064   function To_Bignum (X : Long_Long_Integer) return Bignum is
1065      R : Bignum;
1066
1067   begin
1068      if X = 0 then
1069         R := Allocate_Bignum (0);
1070
1071      --  One word result
1072
1073      elsif X in -(2 ** 32 - 1) .. +(2 ** 32 - 1) then
1074         R := Allocate_Bignum (1);
1075         R.D (1) := SD (abs (X));
1076
1077      --  Largest negative number annoyance
1078
1079      elsif X = Long_Long_Integer'First then
1080         R := Allocate_Bignum (2);
1081         R.D (1) := 2 ** 31;
1082         R.D (2) := 0;
1083
1084      --  Normal two word case
1085
1086      else
1087         R := Allocate_Bignum (2);
1088         R.D (2) := SD (abs (X) mod Base);
1089         R.D (1) := SD (abs (X) / Base);
1090      end if;
1091
1092      R.Neg := X < 0;
1093      return R;
1094   end To_Bignum;
1095
1096end System.Bignums;
1097