PR target/16201
[official-gcc.git] / gcc / ada / s-fatgen.adb
blobac0b0a5f55945d754fbccb01e441b98ef1648418
1 ------------------------------------------------------------------------------
2 -- --
3 -- GNAT COMPILER COMPONENTS --
4 -- --
5 -- S Y S T E M . F A T _ G E N --
6 -- --
7 -- B o d y --
8 -- --
9 -- Copyright (C) 1992-2004 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 2, 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. See the GNU General Public License --
17 -- for more details. You should have received a copy of the GNU General --
18 -- Public License distributed with GNAT; see file COPYING. If not, write --
19 -- to the Free Software Foundation, 59 Temple Place - Suite 330, Boston, --
20 -- MA 02111-1307, USA. --
21 -- --
22 -- As a special exception, if other files instantiate generics from this --
23 -- unit, or you link this unit with other files to produce an executable, --
24 -- this unit does not by itself cause the resulting executable to be --
25 -- covered by the GNU General Public License. This exception does not --
26 -- however invalidate any other reasons why the executable file might be --
27 -- covered by the GNU Public License. --
28 -- --
29 -- GNAT was originally developed by the GNAT team at New York University. --
30 -- Extensive contributions were provided by Ada Core Technologies Inc. --
31 -- --
32 ------------------------------------------------------------------------------
34 -- The implementation here is portable to any IEEE implementation. It does
35 -- not handle non-binary radix, and also assumes that model numbers and
36 -- machine numbers are basically identical, which is not true of all possible
37 -- floating-point implementations. On a non-IEEE machine, this body must be
38 -- specialized appropriately, or better still, its generic instantiations
39 -- should be replaced by efficient machine-specific code.
41 with Ada.Unchecked_Conversion;
42 with System;
43 package body System.Fat_Gen is
45 Float_Radix : constant T := T (T'Machine_Radix);
46 Radix_To_M_Minus_1 : constant T := Float_Radix ** (T'Machine_Mantissa - 1);
48 pragma Assert (T'Machine_Radix = 2);
49 -- This version does not handle radix 16
51 -- Constants for Decompose and Scaling
53 Rad : constant T := T (T'Machine_Radix);
54 Invrad : constant T := 1.0 / Rad;
56 subtype Expbits is Integer range 0 .. 6;
57 -- 2 ** (2 ** 7) might overflow. how big can radix-16 exponents get?
59 Log_Power : constant array (Expbits) of Integer := (1, 2, 4, 8, 16, 32, 64);
61 R_Power : constant array (Expbits) of T :=
62 (Rad ** 1,
63 Rad ** 2,
64 Rad ** 4,
65 Rad ** 8,
66 Rad ** 16,
67 Rad ** 32,
68 Rad ** 64);
70 R_Neg_Power : constant array (Expbits) of T :=
71 (Invrad ** 1,
72 Invrad ** 2,
73 Invrad ** 4,
74 Invrad ** 8,
75 Invrad ** 16,
76 Invrad ** 32,
77 Invrad ** 64);
79 -----------------------
80 -- Local Subprograms --
81 -----------------------
83 procedure Decompose (XX : T; Frac : out T; Expo : out UI);
84 -- Decomposes a floating-point number into fraction and exponent parts.
85 -- Both results are signed, with Frac having the sign of XX, and UI has
86 -- the sign of the exponent. The absolute value of Frac is in the range
87 -- 0.0 <= Frac < 1.0. If Frac = 0.0 or -0.0, then Expo is always zero.
89 function Gradual_Scaling (Adjustment : UI) return T;
90 -- Like Scaling with a first argument of 1.0, but returns the smallest
91 -- denormal rather than zero when the adjustment is smaller than
92 -- Machine_Emin. Used for Succ and Pred.
94 --------------
95 -- Adjacent --
96 --------------
98 function Adjacent (X, Towards : T) return T is
99 begin
100 if Towards = X then
101 return X;
103 elsif Towards > X then
104 return Succ (X);
106 else
107 return Pred (X);
108 end if;
109 end Adjacent;
111 -------------
112 -- Ceiling --
113 -------------
115 function Ceiling (X : T) return T is
116 XT : constant T := Truncation (X);
118 begin
119 if X <= 0.0 then
120 return XT;
122 elsif X = XT then
123 return X;
125 else
126 return XT + 1.0;
127 end if;
128 end Ceiling;
130 -------------
131 -- Compose --
132 -------------
134 function Compose (Fraction : T; Exponent : UI) return T is
135 Arg_Frac : T;
136 Arg_Exp : UI;
137 begin
138 Decompose (Fraction, Arg_Frac, Arg_Exp);
139 return Scaling (Arg_Frac, Exponent);
140 end Compose;
142 ---------------
143 -- Copy_Sign --
144 ---------------
146 function Copy_Sign (Value, Sign : T) return T is
147 Result : T;
149 function Is_Negative (V : T) return Boolean;
150 pragma Import (Intrinsic, Is_Negative);
152 begin
153 Result := abs Value;
155 if Is_Negative (Sign) then
156 return -Result;
157 else
158 return Result;
159 end if;
160 end Copy_Sign;
162 ---------------
163 -- Decompose --
164 ---------------
166 procedure Decompose (XX : T; Frac : out T; Expo : out UI) is
167 X : constant T := T'Machine (XX);
169 begin
170 if X = 0.0 then
171 Frac := X;
172 Expo := 0;
174 -- More useful would be defining Expo to be T'Machine_Emin - 1 or
175 -- T'Machine_Emin - T'Machine_Mantissa, which would preserve
176 -- monotonicity of the exponent function ???
178 -- Check for infinities, transfinites, whatnot.
180 elsif X > T'Safe_Last then
181 Frac := Invrad;
182 Expo := T'Machine_Emax + 1;
184 elsif X < T'Safe_First then
185 Frac := -Invrad;
186 Expo := T'Machine_Emax + 2; -- how many extra negative values?
188 else
189 -- Case of nonzero finite x. Essentially, we just multiply
190 -- by Rad ** (+-2**N) to reduce the range.
192 declare
193 Ax : T := abs X;
194 Ex : UI := 0;
196 -- Ax * Rad ** Ex is invariant.
198 begin
199 if Ax >= 1.0 then
200 while Ax >= R_Power (Expbits'Last) loop
201 Ax := Ax * R_Neg_Power (Expbits'Last);
202 Ex := Ex + Log_Power (Expbits'Last);
203 end loop;
205 -- Ax < Rad ** 64
207 for N in reverse Expbits'First .. Expbits'Last - 1 loop
208 if Ax >= R_Power (N) then
209 Ax := Ax * R_Neg_Power (N);
210 Ex := Ex + Log_Power (N);
211 end if;
213 -- Ax < R_Power (N)
214 end loop;
216 -- 1 <= Ax < Rad
218 Ax := Ax * Invrad;
219 Ex := Ex + 1;
221 else
222 -- 0 < ax < 1
224 while Ax < R_Neg_Power (Expbits'Last) loop
225 Ax := Ax * R_Power (Expbits'Last);
226 Ex := Ex - Log_Power (Expbits'Last);
227 end loop;
229 -- Rad ** -64 <= Ax < 1
231 for N in reverse Expbits'First .. Expbits'Last - 1 loop
232 if Ax < R_Neg_Power (N) then
233 Ax := Ax * R_Power (N);
234 Ex := Ex - Log_Power (N);
235 end if;
237 -- R_Neg_Power (N) <= Ax < 1
238 end loop;
239 end if;
241 if X > 0.0 then
242 Frac := Ax;
243 else
244 Frac := -Ax;
245 end if;
247 Expo := Ex;
248 end;
249 end if;
250 end Decompose;
252 --------------
253 -- Exponent --
254 --------------
256 function Exponent (X : T) return UI is
257 X_Frac : T;
258 X_Exp : UI;
260 begin
261 Decompose (X, X_Frac, X_Exp);
262 return X_Exp;
263 end Exponent;
265 -----------
266 -- Floor --
267 -----------
269 function Floor (X : T) return T is
270 XT : constant T := Truncation (X);
272 begin
273 if X >= 0.0 then
274 return XT;
276 elsif XT = X then
277 return X;
279 else
280 return XT - 1.0;
281 end if;
282 end Floor;
284 --------------
285 -- Fraction --
286 --------------
288 function Fraction (X : T) return T is
289 X_Frac : T;
290 X_Exp : UI;
292 begin
293 Decompose (X, X_Frac, X_Exp);
294 return X_Frac;
295 end Fraction;
297 ---------------------
298 -- Gradual_Scaling --
299 ---------------------
301 function Gradual_Scaling (Adjustment : UI) return T is
302 Y : T;
303 Y1 : T;
304 Ex : UI := Adjustment;
306 begin
307 if Adjustment < T'Machine_Emin - 1 then
308 Y := 2.0 ** T'Machine_Emin;
309 Y1 := Y;
310 Ex := Ex - T'Machine_Emin;
311 while Ex < 0 loop
312 Y := T'Machine (Y / 2.0);
314 if Y = 0.0 then
315 return Y1;
316 end if;
318 Ex := Ex + 1;
319 Y1 := Y;
320 end loop;
322 return Y1;
324 else
325 return Scaling (1.0, Adjustment);
326 end if;
327 end Gradual_Scaling;
329 ------------------
330 -- Leading_Part --
331 ------------------
333 function Leading_Part (X : T; Radix_Digits : UI) return T is
334 L : UI;
335 Y, Z : T;
337 begin
338 if Radix_Digits >= T'Machine_Mantissa then
339 return X;
341 elsif Radix_Digits <= 0 then
342 raise Constraint_Error;
344 else
345 L := Exponent (X) - Radix_Digits;
346 Y := Truncation (Scaling (X, -L));
347 Z := Scaling (Y, L);
348 return Z;
349 end if;
350 end Leading_Part;
352 -------------
353 -- Machine --
354 -------------
356 -- The trick with Machine is to force the compiler to store the result
357 -- in memory so that we do not have extra precision used. The compiler
358 -- is clever, so we have to outwit its possible optimizations! We do
359 -- this by using an intermediate pragma Volatile location.
361 function Machine (X : T) return T is
362 Temp : T;
363 pragma Volatile (Temp);
364 begin
365 Temp := X;
366 return Temp;
367 end Machine;
369 -----------
370 -- Model --
371 -----------
373 -- We treat Model as identical to Machine. This is true of IEEE and other
374 -- nice floating-point systems, but not necessarily true of all systems.
376 function Model (X : T) return T is
377 begin
378 return Machine (X);
379 end Model;
381 ----------
382 -- Pred --
383 ----------
385 -- Subtract from the given number a number equivalent to the value of its
386 -- least significant bit. Given that the most significant bit represents
387 -- a value of 1.0 * radix ** (exp - 1), the value we want is obtained by
388 -- shifting this by (mantissa-1) bits to the right, i.e. decreasing the
389 -- exponent by that amount.
391 -- Zero has to be treated specially, since its exponent is zero
393 function Pred (X : T) return T is
394 X_Frac : T;
395 X_Exp : UI;
397 begin
398 if X = 0.0 then
399 return -Succ (X);
401 else
402 Decompose (X, X_Frac, X_Exp);
404 -- A special case, if the number we had was a positive power of
405 -- two, then we want to subtract half of what we would otherwise
406 -- subtract, since the exponent is going to be reduced.
408 -- Note that X_Frac has the same sign as X, so if X_Frac is 0.5,
409 -- then we know that we have a positive number (and hence a
410 -- positive power of 2).
412 if X_Frac = 0.5 then
413 return X - Gradual_Scaling (X_Exp - T'Machine_Mantissa - 1);
415 -- Otherwise the exponent is unchanged
417 else
418 return X - Gradual_Scaling (X_Exp - T'Machine_Mantissa);
419 end if;
420 end if;
421 end Pred;
423 ---------------
424 -- Remainder --
425 ---------------
427 function Remainder (X, Y : T) return T is
428 A : T;
429 B : T;
430 Arg : T;
431 P : T;
432 Arg_Frac : T;
433 P_Frac : T;
434 Sign_X : T;
435 IEEE_Rem : T;
436 Arg_Exp : UI;
437 P_Exp : UI;
438 K : UI;
439 P_Even : Boolean;
441 begin
442 if Y = 0.0 then
443 raise Constraint_Error;
444 end if;
446 if X > 0.0 then
447 Sign_X := 1.0;
448 Arg := X;
449 else
450 Sign_X := -1.0;
451 Arg := -X;
452 end if;
454 P := abs Y;
456 if Arg < P then
457 P_Even := True;
458 IEEE_Rem := Arg;
459 P_Exp := Exponent (P);
461 else
462 Decompose (Arg, Arg_Frac, Arg_Exp);
463 Decompose (P, P_Frac, P_Exp);
465 P := Compose (P_Frac, Arg_Exp);
466 K := Arg_Exp - P_Exp;
467 P_Even := True;
468 IEEE_Rem := Arg;
470 for Cnt in reverse 0 .. K loop
471 if IEEE_Rem >= P then
472 P_Even := False;
473 IEEE_Rem := IEEE_Rem - P;
474 else
475 P_Even := True;
476 end if;
478 P := P * 0.5;
479 end loop;
480 end if;
482 -- That completes the calculation of modulus remainder. The final
483 -- step is get the IEEE remainder. Here we need to compare Rem with
484 -- (abs Y) / 2. We must be careful of unrepresentable Y/2 value
485 -- caused by subnormal numbers
487 if P_Exp >= 0 then
488 A := IEEE_Rem;
489 B := abs Y * 0.5;
491 else
492 A := IEEE_Rem * 2.0;
493 B := abs Y;
494 end if;
496 if A > B or else (A = B and then not P_Even) then
497 IEEE_Rem := IEEE_Rem - abs Y;
498 end if;
500 return Sign_X * IEEE_Rem;
501 end Remainder;
503 --------------
504 -- Rounding --
505 --------------
507 function Rounding (X : T) return T is
508 Result : T;
509 Tail : T;
511 begin
512 Result := Truncation (abs X);
513 Tail := abs X - Result;
515 if Tail >= 0.5 then
516 Result := Result + 1.0;
517 end if;
519 if X > 0.0 then
520 return Result;
522 elsif X < 0.0 then
523 return -Result;
525 -- For zero case, make sure sign of zero is preserved
527 else
528 return X;
529 end if;
530 end Rounding;
532 -------------
533 -- Scaling --
534 -------------
536 -- Return x * rad ** adjustment quickly,
537 -- or quietly underflow to zero, or overflow naturally.
539 function Scaling (X : T; Adjustment : UI) return T is
540 begin
541 if X = 0.0 or else Adjustment = 0 then
542 return X;
543 end if;
545 -- Nonzero x. essentially, just multiply repeatedly by Rad ** (+-2**n).
547 declare
548 Y : T := X;
549 Ex : UI := Adjustment;
551 -- Y * Rad ** Ex is invariant
553 begin
554 if Ex < 0 then
555 while Ex <= -Log_Power (Expbits'Last) loop
556 Y := Y * R_Neg_Power (Expbits'Last);
557 Ex := Ex + Log_Power (Expbits'Last);
558 end loop;
560 -- -64 < Ex <= 0
562 for N in reverse Expbits'First .. Expbits'Last - 1 loop
563 if Ex <= -Log_Power (N) then
564 Y := Y * R_Neg_Power (N);
565 Ex := Ex + Log_Power (N);
566 end if;
568 -- -Log_Power (N) < Ex <= 0
569 end loop;
571 -- Ex = 0
573 else
574 -- Ex >= 0
576 while Ex >= Log_Power (Expbits'Last) loop
577 Y := Y * R_Power (Expbits'Last);
578 Ex := Ex - Log_Power (Expbits'Last);
579 end loop;
581 -- 0 <= Ex < 64
583 for N in reverse Expbits'First .. Expbits'Last - 1 loop
584 if Ex >= Log_Power (N) then
585 Y := Y * R_Power (N);
586 Ex := Ex - Log_Power (N);
587 end if;
589 -- 0 <= Ex < Log_Power (N)
590 end loop;
592 -- Ex = 0
593 end if;
595 return Y;
596 end;
597 end Scaling;
599 ----------
600 -- Succ --
601 ----------
603 -- Similar computation to that of Pred: find value of least significant
604 -- bit of given number, and add. Zero has to be treated specially since
605 -- the exponent can be zero, and also we want the smallest denormal if
606 -- denormals are supported.
608 function Succ (X : T) return T is
609 X_Frac : T;
610 X_Exp : UI;
611 X1, X2 : T;
613 begin
614 if X = 0.0 then
615 X1 := 2.0 ** T'Machine_Emin;
617 -- Following loop generates smallest denormal
619 loop
620 X2 := T'Machine (X1 / 2.0);
621 exit when X2 = 0.0;
622 X1 := X2;
623 end loop;
625 return X1;
627 else
628 Decompose (X, X_Frac, X_Exp);
630 -- A special case, if the number we had was a negative power of
631 -- two, then we want to add half of what we would otherwise add,
632 -- since the exponent is going to be reduced.
634 -- Note that X_Frac has the same sign as X, so if X_Frac is -0.5,
635 -- then we know that we have a ngeative number (and hence a
636 -- negative power of 2).
638 if X_Frac = -0.5 then
639 return X + Gradual_Scaling (X_Exp - T'Machine_Mantissa - 1);
641 -- Otherwise the exponent is unchanged
643 else
644 return X + Gradual_Scaling (X_Exp - T'Machine_Mantissa);
645 end if;
646 end if;
647 end Succ;
649 ----------------
650 -- Truncation --
651 ----------------
653 -- The basic approach is to compute
655 -- T'Machine (RM1 + N) - RM1.
657 -- where N >= 0.0 and RM1 = radix ** (mantissa - 1)
659 -- This works provided that the intermediate result (RM1 + N) does not
660 -- have extra precision (which is why we call Machine). When we compute
661 -- RM1 + N, the exponent of N will be normalized and the mantissa shifted
662 -- shifted appropriately so the lower order bits, which cannot contribute
663 -- to the integer part of N, fall off on the right. When we subtract RM1
664 -- again, the significant bits of N are shifted to the left, and what we
665 -- have is an integer, because only the first e bits are different from
666 -- zero (assuming binary radix here).
668 function Truncation (X : T) return T is
669 Result : T;
671 begin
672 Result := abs X;
674 if Result >= Radix_To_M_Minus_1 then
675 return Machine (X);
677 else
678 Result := Machine (Radix_To_M_Minus_1 + Result) - Radix_To_M_Minus_1;
680 if Result > abs X then
681 Result := Result - 1.0;
682 end if;
684 if X > 0.0 then
685 return Result;
687 elsif X < 0.0 then
688 return -Result;
690 -- For zero case, make sure sign of zero is preserved
692 else
693 return X;
694 end if;
695 end if;
697 end Truncation;
699 -----------------------
700 -- Unbiased_Rounding --
701 -----------------------
703 function Unbiased_Rounding (X : T) return T is
704 Abs_X : constant T := abs X;
705 Result : T;
706 Tail : T;
708 begin
709 Result := Truncation (Abs_X);
710 Tail := Abs_X - Result;
712 if Tail > 0.5 then
713 Result := Result + 1.0;
715 elsif Tail = 0.5 then
716 Result := 2.0 * Truncation ((Result / 2.0) + 0.5);
717 end if;
719 if X > 0.0 then
720 return Result;
722 elsif X < 0.0 then
723 return -Result;
725 -- For zero case, make sure sign of zero is preserved
727 else
728 return X;
729 end if;
731 end Unbiased_Rounding;
733 -----------
734 -- Valid --
735 -----------
737 function Valid (X : access T) return Boolean is
739 IEEE_Emin : constant Integer := T'Machine_Emin - 1;
740 IEEE_Emax : constant Integer := T'Machine_Emax - 1;
742 IEEE_Bias : constant Integer := -(IEEE_Emin - 1);
744 subtype IEEE_Exponent_Range is
745 Integer range IEEE_Emin - 1 .. IEEE_Emax + 1;
747 -- The implementation of this floating point attribute uses
748 -- a representation type Float_Rep that allows direct access to
749 -- the exponent and mantissa parts of a floating point number.
751 -- The Float_Rep type is an array of Float_Word elements. This
752 -- representation is chosen to make it possible to size the
753 -- type based on a generic parameter. Since the array size is
754 -- known at compile-time, efficient code can still be generated.
755 -- The size of Float_Word elements should be large enough to allow
756 -- accessing the exponent in one read, but small enough so that all
757 -- floating point object sizes are a multiple of the Float_Word'Size.
759 -- The following conditions must be met for all possible
760 -- instantiations of the attributes package:
762 -- - T'Size is an integral multiple of Float_Word'Size
764 -- - The exponent and sign are completely contained in a single
765 -- component of Float_Rep, named Most_Significant_Word (MSW).
767 -- - The sign occupies the most significant bit of the MSW
768 -- and the exponent is in the following bits.
769 -- Unused bits (if any) are in the least significant part.
771 type Float_Word is mod 2**Positive'Min (System.Word_Size, 32);
772 type Rep_Index is range 0 .. 7;
774 Rep_Last : constant Rep_Index := (T'Size - 1) / Float_Word'Size;
776 type Float_Rep is array (Rep_Index range 0 .. Rep_Last) of Float_Word;
778 pragma Suppress_Initialization (Float_Rep);
779 -- This pragma supresses the generation of an initialization procedure
780 -- for type Float_Rep when operating in Initialize/Normalize_Scalars
781 -- mode. This is not just a matter of efficiency, but of functionality,
782 -- since Valid has a pragma Inline_Always, which is not permitted if
783 -- there are nested subprograms present.
785 Most_Significant_Word : constant Rep_Index :=
786 Rep_Last * Standard'Default_Bit_Order;
787 -- Finding the location of the Exponent_Word is a bit tricky.
788 -- In general we assume Word_Order = Bit_Order.
789 -- This expression needs to be refined for VMS.
791 Exponent_Factor : constant Float_Word :=
792 2**(Float_Word'Size - 1) /
793 Float_Word (IEEE_Emax - IEEE_Emin + 3) *
794 Boolean'Pos (T'Size /= 96) +
795 Boolean'Pos (T'Size = 96);
796 -- Factor that the extracted exponent needs to be divided by
797 -- to be in range 0 .. IEEE_Emax - IEEE_Emin + 2.
798 -- Special kludge: Exponent_Factor is 0 for x86 double extended
799 -- as GCC adds 16 unused bits to the type.
801 Exponent_Mask : constant Float_Word :=
802 Float_Word (IEEE_Emax - IEEE_Emin + 2) *
803 Exponent_Factor;
804 -- Value needed to mask out the exponent field.
805 -- This assumes that the range IEEE_Emin - 1 .. IEEE_Emax + 1
806 -- contains 2**N values, for some N in Natural.
808 function To_Float is new Ada.Unchecked_Conversion (Float_Rep, T);
810 type Float_Access is access all T;
811 function To_Address is
812 new Ada.Unchecked_Conversion (Float_Access, System.Address);
814 XA : constant System.Address := To_Address (Float_Access (X));
816 R : Float_Rep;
817 pragma Import (Ada, R);
818 for R'Address use XA;
819 -- R is a view of the input floating-point parameter. Note that we
820 -- must avoid copying the actual bits of this parameter in float
821 -- form (since it may be a signalling NaN.
823 E : constant IEEE_Exponent_Range :=
824 Integer ((R (Most_Significant_Word) and Exponent_Mask) /
825 Exponent_Factor)
826 - IEEE_Bias;
827 -- Mask/Shift T to only get bits from the exponent
828 -- Then convert biased value to integer value.
830 SR : Float_Rep;
831 -- Float_Rep representation of significant of X.all
833 begin
834 if T'Denorm then
836 -- All denormalized numbers are valid, so only invalid numbers
837 -- are overflows and NaN's, both with exponent = Emax + 1.
839 return E /= IEEE_Emax + 1;
841 end if;
843 -- All denormalized numbers except 0.0 are invalid
845 -- Set exponent of X to zero, so we end up with the significand, which
846 -- definitely is a valid number and can be converted back to a float.
848 SR := R;
849 SR (Most_Significant_Word) :=
850 (SR (Most_Significant_Word)
851 and not Exponent_Mask) + Float_Word (IEEE_Bias) * Exponent_Factor;
853 return (E in IEEE_Emin .. IEEE_Emax) or else
854 ((E = IEEE_Emin - 1) and then abs To_Float (SR) = 1.0);
855 end Valid;
857 ---------------------
858 -- Unaligned_Valid --
859 ---------------------
861 function Unaligned_Valid (A : System.Address) return Boolean is
862 subtype FS is String (1 .. T'Size / Character'Size);
863 type FSP is access FS;
865 function To_FSP is new Ada.Unchecked_Conversion (Address, FSP);
867 Local_T : aliased T;
869 begin
870 To_FSP (Local_T'Address).all := To_FSP (A).all;
871 return Valid (Local_T'Access);
872 end Unaligned_Valid;
874 end System.Fat_Gen;