
git-svn-id: https://svn.code.sf.net/p/lazarus-ccr/svn@6159 8e941d3f-bd1b-0410-a28a-d453659cc2b4
1309 lines
34 KiB
ObjectPascal
1309 lines
34 KiB
ObjectPascal
// Upgraded to Delphi 2009: Sebastian Zierer
|
|
|
|
(* ***** BEGIN LICENSE BLOCK *****
|
|
* Version: MPL 1.1
|
|
*
|
|
* The contents of this file are subject to the Mozilla Public License Version
|
|
* 1.1 (the "License"); you may not use this file except in compliance with
|
|
* the License. You may obtain a copy of the License at
|
|
* http://www.mozilla.org/MPL/
|
|
*
|
|
* Software distributed under the License is distributed on an "AS IS" basis,
|
|
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
|
|
* for the specific language governing rights and limitations under the
|
|
* License.
|
|
*
|
|
* The Original Code is TurboPower SysTools
|
|
*
|
|
* The Initial Developer of the Original Code is
|
|
* TurboPower Software
|
|
*
|
|
* Portions created by the Initial Developer are Copyright (C) 1996-2002
|
|
* the Initial Developer. All Rights Reserved.
|
|
*
|
|
* Contributor(s):
|
|
*
|
|
* ***** END LICENSE BLOCK ***** *)
|
|
|
|
{*********************************************************}
|
|
{* SysTools: StDecMth.pas 4.04 *}
|
|
{*********************************************************}
|
|
{* SysTools: Class for doing decimal arithmetic *}
|
|
{*********************************************************}
|
|
|
|
{$IFDEF FPC}
|
|
{$mode DELPHI}
|
|
{$ENDIF}
|
|
|
|
//{$I StDefine.inc}
|
|
|
|
unit StDecMth;
|
|
|
|
interface
|
|
|
|
{Note: StDecMth declares and implements TStDecimal. This is a fixed-
|
|
point value with a total of 38 significant digits of which
|
|
16 are to the right of the decimal point.}
|
|
|
|
uses
|
|
SysUtils;
|
|
|
|
type
|
|
TStRoundMethod = ( {different rounding methods...}
|
|
rmNormal, {..normal (round away from zero if half way)}
|
|
rmTrunc, {..truncate (always round to zero)}
|
|
rmBankers, {..bankers (round to even digit if half way)}
|
|
rmUp); {..force round up (always round from zero)}
|
|
|
|
TStInt128 = array [0..3] of longint; // must be longint, not DWORD
|
|
|
|
TStDecimal = class
|
|
private
|
|
FInt : TStInt128;
|
|
protected
|
|
function dcGetAsStr : AnsiString;
|
|
procedure dcSetFromStr(const aValue : AnsiString); {!!.02}
|
|
public
|
|
constructor Create;
|
|
destructor Destroy; override;
|
|
|
|
function Compare(X : TStDecimal) : integer;
|
|
{-returns <0 if Self < X, 0 is equal, >0 otherwise}
|
|
function IsNegative : boolean;
|
|
{-returns Self < 0.0}
|
|
function IsOne : boolean;
|
|
{-returns Self = 1.0}
|
|
function IsPositive : boolean;
|
|
{-returns Self > 0.0}
|
|
function IsZero : boolean;
|
|
{-returns Self = 0.0}
|
|
procedure SetToOne;
|
|
{-sets Self equal to 1.0}
|
|
procedure SetToZero;
|
|
{-sets Self equal to 0.0}
|
|
|
|
procedure Assign(X : TStDecimal);
|
|
{-sets Self equal to X}
|
|
procedure AssignFromFloat(aValue : double);
|
|
{-sets Self equal to aValue}
|
|
procedure AssignFromInt(aValue : integer);
|
|
{-sets Self equal to aValue}
|
|
|
|
function AsFloat : double;
|
|
{-returns Self as an floating point value}
|
|
function AsInt(aRound : TStRoundMethod) : integer;
|
|
{-returns Self as an integer, rounded}
|
|
|
|
procedure Abs;
|
|
{-calculates Self := Abs(Self)}
|
|
procedure Add(X : TStDecimal);
|
|
{-calculates Self := Self + X}
|
|
procedure AddOne;
|
|
{-calculates Self := Self + 1.0}
|
|
procedure ChangeSign;
|
|
{-calculates Self := ChgSign(Self)}
|
|
procedure Divide(X : TStDecimal);
|
|
{-calculates Self := Self div X}
|
|
procedure Multiply(X : TStDecimal);
|
|
{-calculates Self := Self * X}
|
|
procedure RaiseToPower(N : integer);
|
|
{-calculates Self := Self ^ N}
|
|
procedure Round(aRound : TStRoundMethod; aDecPl : integer);
|
|
{-calculates Self := Round(Self)}
|
|
procedure Subtract(X : TStDecimal);
|
|
{-calculates Self := Self - X}
|
|
procedure SubtractOne;
|
|
{-calculates Self := Self - 1}
|
|
|
|
property AsString : AnsiString read dcGetAsStr write dcSetFromStr;
|
|
{-returns Self as a string, sets Self from a string}
|
|
end;
|
|
|
|
implementation
|
|
|
|
uses
|
|
StConst,
|
|
StBase;
|
|
|
|
type
|
|
TStInt256 = array [0..7] of integer;
|
|
TStInt192 = array [0..5] of integer;
|
|
|
|
const
|
|
MaxDecPl = 16;
|
|
|
|
Int128One_0 = longint($6FC10000);
|
|
Int128One_1 = longint($002386F2);
|
|
|
|
PowerOf10 : array [0..MaxDecPl div 2] of integer =
|
|
(1, 10, 100, 1000, 10000, 100000, 1000000, 10000000,
|
|
100000000);
|
|
|
|
{===Helper routines==================================================}
|
|
procedure Int256Div10E8(var X : TStInt256; var aRem : integer);
|
|
{Note: this routine assumes X is positive}
|
|
asm
|
|
push ebx // save ebx
|
|
|
|
push edx // save address of remainder variable
|
|
|
|
mov ecx, 100000000 // we're dividing by 10^8
|
|
mov ebx, eax // ebx points to X
|
|
|
|
xor edx, edx // start off with high dividend digit zero
|
|
mov eax, [ebx+28] // get last 32-bit digit
|
|
div ecx // divide by 10: eax is quotient, edx remainder
|
|
mov [ebx+28], eax // save highest quotient digit
|
|
|
|
mov eax, [ebx+24] // get next 32-bit digit
|
|
div ecx // divide by 10: eax is quotient, edx remainder
|
|
mov [ebx+24], eax // save next quotient digit
|
|
|
|
mov eax, [ebx+20] // get next 32-bit digit
|
|
div ecx // divide by 10: eax is quotient, edx remainder
|
|
mov [ebx+20], eax // save next quotient digit
|
|
|
|
mov eax, [ebx+16] // get next 32-bit digit
|
|
div ecx // divide by 10: eax is quotient, edx remainder
|
|
mov [ebx+16], eax // save next quotient digit
|
|
|
|
mov eax, [ebx+12] // get next 32-bit digit
|
|
div ecx // divide by 10: eax is quotient, edx remainder
|
|
mov [ebx+12], eax // save next quotient digit
|
|
|
|
mov eax, [ebx+8] // get next 32-bit digit
|
|
div ecx // divide by 10: eax is quotient, edx remainder
|
|
mov [ebx+8], eax // save next quotient digit
|
|
|
|
mov eax, [ebx+4] // get next 32-bit digit
|
|
div ecx // divide by 10: eax is quotient, edx remainder
|
|
mov [ebx+4], eax // save next quotient digit
|
|
|
|
mov eax, [ebx] // get first 32-bit digit
|
|
div ecx // divide by 10: eax is quotient, edx remainder
|
|
mov [ebx], eax // save first quotient digit
|
|
|
|
pop eax // pop off the address of remainder variable
|
|
mov [eax], edx // store remainder
|
|
|
|
pop ebx // restore ebx
|
|
end;
|
|
{--------}
|
|
procedure Int192Times10E8(var X : TStInt192);
|
|
{Note: this routine assumes X is positive}
|
|
asm
|
|
push ebx // save ebx
|
|
push ebp // save ebp
|
|
|
|
mov ecx, 100000000 // we're multiplying by 10^8
|
|
mov ebx, eax // ebx points to X
|
|
|
|
mov eax, [ebx] // get the first 32-bit digit
|
|
mul ecx // multiply it by 10^8 to give answer in edx:eax
|
|
mov [ebx], eax // save first digit of result
|
|
mov ebp, edx // save overflow
|
|
|
|
mov eax, [ebx+4] // get the second 32-bit digit
|
|
mul ecx // multiply it by 10^8 to give answer in edx:eax
|
|
add eax, ebp // add the overflow from the first digit
|
|
adc edx, 0
|
|
mov [ebx+4], eax // save second digit of result
|
|
mov ebp, edx // save overflow
|
|
|
|
mov eax, [ebx+8] // get the third 32-bit digit
|
|
mul ecx // multiply it by 10^8 to give answer in edx:eax
|
|
add eax, ebp // add the overflow from the second digit
|
|
adc edx, 0
|
|
mov [ebx+8], eax // save third digit of result
|
|
mov ebp, edx // save overflow
|
|
|
|
mov eax, [ebx+12] // get the fourth 32-bit digit
|
|
mul ecx // multiply it by 10^8 to give answer in edx:eax
|
|
add eax, ebp // add the overflow from the third digit
|
|
adc edx, 0
|
|
mov [ebx+12], eax // save fourth digit of result
|
|
mov ebp, edx // save overflow
|
|
|
|
mov eax, [ebx+16] // get the fifth 32-bit digit
|
|
mul ecx // multiply it by 10^8 to give answer in edx:eax
|
|
add eax, ebp // add the overflow from the fourth digit
|
|
adc edx, 0
|
|
mov [ebx+16], eax // save fifth digit of result
|
|
mov ebp, edx // save overflow
|
|
|
|
mov eax, [ebx+20] // get the sixth 32-bit digit
|
|
mul ecx // multiply it by 10^8 to give answer in edx:eax
|
|
add eax, ebp // add the overflow from the fifth digit
|
|
mov [ebx+20], eax // save sixth digit of result
|
|
|
|
pop ebp // restore ebp
|
|
pop ebx // restore ebx
|
|
end;
|
|
{--------}
|
|
function Int32MultPrim(X, Y : longint;
|
|
var P : longint; Carry : longint) : longint;
|
|
asm
|
|
{Note: calculates X * Y + P + Carry
|
|
returns answer in P, with overflow as result value}
|
|
mul edx
|
|
add eax, [ecx]
|
|
adc edx, 0
|
|
add eax, Carry
|
|
adc edx, 0
|
|
mov [ecx], eax
|
|
mov eax, edx
|
|
end;
|
|
{--------}
|
|
procedure Int128Add(var X : TStInt128; const Y : TStInt128);
|
|
asm
|
|
push ebx
|
|
mov ecx, [edx]
|
|
mov ebx, [edx+4]
|
|
add [eax], ecx
|
|
adc [eax+4], ebx
|
|
mov ecx, [edx+8]
|
|
mov ebx, [edx+12]
|
|
adc [eax+8], ecx
|
|
adc [eax+12], ebx
|
|
pop ebx
|
|
end;
|
|
{--------}
|
|
procedure Int128AddInt(var X : TStInt128; aDigit : integer);
|
|
asm
|
|
add [eax], edx
|
|
adc dword ptr [eax+4], 0
|
|
adc dword ptr [eax+8], 0
|
|
adc dword ptr [eax+12], 0
|
|
end;
|
|
{--------}
|
|
procedure Int128ChgSign(var X : TStInt128);
|
|
asm
|
|
mov ecx, [eax]
|
|
mov edx, [eax+4]
|
|
not ecx
|
|
not edx
|
|
add ecx, 1
|
|
adc edx, 0
|
|
mov [eax], ecx
|
|
mov [eax+4], edx
|
|
mov ecx, [eax+8]
|
|
mov edx, [eax+12]
|
|
not ecx
|
|
not edx
|
|
adc ecx, 0
|
|
adc edx, 0
|
|
mov [eax+8], ecx
|
|
mov [eax+12], edx
|
|
end;
|
|
{--------}
|
|
function Int128Compare(const X, Y : TStInt128) : integer;
|
|
asm
|
|
// Can be called from pascal
|
|
// All registers are preserved, except eax, which returns the
|
|
// result of the comparison
|
|
push ebx
|
|
push ecx
|
|
mov ecx, [eax+12]
|
|
mov ebx, [edx+12]
|
|
xor ecx, $80000000
|
|
xor ebx, $80000000
|
|
cmp ecx, ebx
|
|
jb @@LessThan
|
|
ja @@GreaterThan
|
|
mov ecx, [eax+8]
|
|
mov ebx, [edx+8]
|
|
cmp ecx, ebx
|
|
jb @@LessThan
|
|
ja @@GreaterThan
|
|
mov ecx, [eax+4]
|
|
mov ebx, [edx+4]
|
|
cmp ecx, ebx
|
|
jb @@LessThan
|
|
ja @@GreaterThan
|
|
mov ecx, [eax]
|
|
mov ebx, [edx]
|
|
cmp ecx, ebx
|
|
jb @@LessThan
|
|
ja @@GreaterThan
|
|
xor eax, eax
|
|
jmp @@Exit
|
|
@@LessThan:
|
|
mov eax, -1
|
|
jmp @@Exit
|
|
@@GreaterThan:
|
|
mov eax, 1
|
|
@@Exit:
|
|
pop ecx
|
|
pop ebx
|
|
end;
|
|
{--------}
|
|
procedure Int192SHL(var X : TStInt192);
|
|
asm
|
|
// DO NOT CALL FROM PASCAL
|
|
// IN: eax -> 192-bit integer to shift left
|
|
// OUT: eax -> 192-bit integer shifted left
|
|
// CF = most significant bit shifted out
|
|
// All registers are preserved
|
|
push ebx
|
|
push ecx
|
|
mov ebx, [eax]
|
|
mov ecx, [eax+4]
|
|
shl ebx, 1
|
|
rcl ecx, 1
|
|
mov [eax], ebx
|
|
mov [eax+4], ecx
|
|
mov ebx, [eax+8]
|
|
mov ecx, [eax+12]
|
|
rcl ebx, 1
|
|
rcl ecx, 1
|
|
mov [eax+8], ebx
|
|
mov [eax+12], ecx
|
|
mov ebx, [eax+16]
|
|
mov ecx, [eax+20]
|
|
rcl ebx, 1
|
|
rcl ecx, 1
|
|
mov [eax+16], ebx
|
|
mov [eax+20], ecx
|
|
pop ecx
|
|
pop ebx
|
|
end;
|
|
{--------}
|
|
procedure Int128RCL(var X : TStInt128);
|
|
asm
|
|
// DO NOT CALL FROM PASCAL
|
|
// IN: eax -> 128-bit integer to shift left
|
|
// CF = least significant bit to shift in
|
|
// OUT: eax -> 128-bit integer shifted left
|
|
// CF -> topmost bit shifted out
|
|
// All registers are preserved
|
|
push ebx
|
|
push ecx
|
|
mov ebx, [eax]
|
|
mov ecx, [eax+4]
|
|
rcl ebx, 1
|
|
rcl ecx, 1
|
|
mov [eax], ebx
|
|
mov [eax+4], ecx
|
|
mov ebx, [eax+8]
|
|
mov ecx, [eax+12]
|
|
rcl ebx, 1
|
|
rcl ecx, 1
|
|
mov [eax+8], ebx
|
|
mov [eax+12], ecx
|
|
pop ecx
|
|
pop ebx
|
|
end;
|
|
{--------}
|
|
procedure Int128FastDivide(var X : TStInt192;
|
|
var Y, aRem : TStInt128);
|
|
asm
|
|
push ebp
|
|
push ebx
|
|
push edi
|
|
push esi
|
|
|
|
mov esi, eax // esi -> dividend
|
|
mov edi, edx // edi -> divisor
|
|
mov ebp, ecx // ebp -> remainder
|
|
|
|
mov ecx, 192 // we'll do the loop for all 192 bits in the
|
|
// dividend
|
|
|
|
xor eax, eax // zero the remainder
|
|
mov [ebp], eax
|
|
mov [ebp+4], eax
|
|
mov [ebp+8], eax
|
|
mov [ebp+12], eax
|
|
|
|
@@GetNextBit:
|
|
mov eax, esi // shift the dividend left, and...
|
|
call Int192SHL
|
|
mov eax, ebp // ...shift the topmost bit into the remainder
|
|
call Int128RCL
|
|
|
|
mov eax, ebp // compare the remainder with the divisor
|
|
mov edx, edi
|
|
call Int128Compare
|
|
|
|
cmp eax, 0 // if the remainder is smaller, we can't
|
|
jl @@TooSmall // subtract the divisor
|
|
|
|
// essentially we've shown that the divisor
|
|
// divides the remainder exactly once, so
|
|
|
|
add dword ptr [esi], 1 // add one to the quotient
|
|
|
|
mov eax, [ebp] // subtract the divisor from the remainder
|
|
mov ebx, [ebp+4]
|
|
sub eax, [edi]
|
|
sbb ebx, [edi+4]
|
|
mov [ebp], eax
|
|
mov [ebp+4], ebx
|
|
mov eax, [ebp+8]
|
|
mov ebx, [ebp+12]
|
|
sbb eax, [edi+8]
|
|
sbb ebx, [edi+12]
|
|
mov [ebp+8], eax
|
|
mov [ebp+12], ebx
|
|
|
|
@@TooSmall:
|
|
dec ecx // go get the next bit to work on
|
|
jnz @@GetNextBit
|
|
|
|
pop esi
|
|
pop edi
|
|
pop ebx
|
|
pop ebp
|
|
end;
|
|
{--------}
|
|
function Int128DivInt(var X : TStInt128; aDivisor : integer) : integer;
|
|
{Note: this routine assumes X is positive}
|
|
asm
|
|
push ebx // save ebx
|
|
|
|
mov ecx, edx // ecx is now the divisor
|
|
mov ebx, eax // ebx points to X
|
|
|
|
xor edx, edx // start off with high dividend digit zero
|
|
mov eax, [ebx+12] // get last 32-bit digit
|
|
div ecx // divide by ecx: eax is quotient, edx remainder
|
|
mov [ebx+12], eax // save highest quotient digit
|
|
|
|
mov eax, [ebx+8] // get next 32-bit digit
|
|
div ecx // divide by ecx: eax is quotient, edx remainder
|
|
mov [ebx+8], eax // save next quotient digit
|
|
|
|
mov eax, [ebx+4] // get next 32-bit digit
|
|
div ecx // divide by ecx: eax is quotient, edx remainder
|
|
mov [ebx+4], eax // save next quotient digit
|
|
|
|
mov eax, [ebx] // get first 32-bit digit
|
|
div ecx // divide by ecx: eax is quotient, edx remainder
|
|
mov [ebx], eax // save first quotient digit
|
|
|
|
mov eax, edx // return remainder
|
|
|
|
pop ebx // restore ebx
|
|
end;
|
|
{--------}
|
|
procedure Int128Divide(var X, Y : TStInt128);
|
|
var
|
|
XTemp : TStInt192;
|
|
Rem : TStInt128;
|
|
begin
|
|
{note: the easy cases have been dealt with
|
|
X and Y are both positive
|
|
X will be set to the quotient X/Y and Y will be trashed}
|
|
|
|
{we need to increase the number of decimal places to 32, so convert
|
|
the 128 bit dividend to a 192 bit one and multiply by 10^16}
|
|
XTemp[0] := X[0];
|
|
XTemp[1] := X[1];
|
|
XTemp[2] := X[2];
|
|
XTemp[3] := X[3];
|
|
XTemp[4] := 0;
|
|
XTemp[5] := 0;
|
|
Int192Times10E8(XTemp);
|
|
Int192Times10E8(XTemp);
|
|
|
|
{Note: this algorithm follows that described by Knuth in volume 2 of
|
|
The Art of Computer Programming. Algorithm D of section 4.3
|
|
as applied to binary numbers (radix=2)}
|
|
|
|
{divide the 192-bit dividend by the 128-bit divisor}
|
|
Int128FastDivide(XTemp, Y, Rem);
|
|
|
|
{have we overflowed? ie, have we divided a very big number by one
|
|
much less than zero}
|
|
if (XTemp[3] < 0) or (XTemp[4] <> 0) or (XTemp[5] <> 0) then
|
|
raise EStDecMathError.Create(stscDecMathDivOverflowS);
|
|
|
|
{return the result of the computation}
|
|
X[0] := XTemp[0];
|
|
X[1] := XTemp[1];
|
|
X[2] := XTemp[2];
|
|
X[3] := XTemp[3];
|
|
end;
|
|
{--------}
|
|
procedure Int128Multiply(var X, Y : TStInt128);
|
|
var
|
|
P : TStInt256;
|
|
XIsNeg : boolean;
|
|
YIsNeg : boolean;
|
|
YInx : integer;
|
|
YDigit : integer;
|
|
Carry : integer;
|
|
YTemp : TStInt128;
|
|
begin
|
|
{Note: calculates X * Y and puts the answer in X}
|
|
|
|
{get rid of the easy cases where one of the operands is zero}
|
|
if (X[0] = 0) and (X[1] = 0) and (X[2] = 0) and (X[3] = 0) then
|
|
Exit;
|
|
if (Y[0] = 0) and (Y[1] = 0) and (Y[2] = 0) and (Y[3] = 0) then begin
|
|
X[0] := 0;
|
|
X[1] := 0;
|
|
X[2] := 0;
|
|
X[3] := 0;
|
|
Exit;
|
|
end;
|
|
|
|
{we might need to trash Y, so we use a local variable}
|
|
YTemp[0] := Y[0];
|
|
YTemp[1] := Y[1];
|
|
YTemp[2] := Y[2];
|
|
YTemp[3] := Y[3];
|
|
|
|
{convert both operands to positive values: we'll fix the sign later}
|
|
XIsNeg := X[3] < 0;
|
|
if XIsNeg then
|
|
Int128ChgSign(X);
|
|
YIsNeg := YTemp[3] < 0;
|
|
if YIsNeg then
|
|
Int128ChgSign(YTemp);
|
|
|
|
{initialize the temporary product}
|
|
P[0] := 0;
|
|
P[1] := 0;
|
|
P[2] := 0;
|
|
P[3] := 0;
|
|
P[4] := 0;
|
|
P[5] := 0;
|
|
P[6] := 0;
|
|
P[7] := 0;
|
|
|
|
{for every digit in Y we shall multiply by all the X digits and sum}
|
|
for YInx := 0 to 3 do begin
|
|
|
|
{get the Y digit}
|
|
YDigit := YTemp[YInx];
|
|
|
|
{there's only something to do if the Y digit is non-zero}
|
|
if (YDigit <> 0) then begin
|
|
|
|
{multiply this digit with all the X digits, storing the result
|
|
in the temporary product}
|
|
Carry := Int32MultPrim(X[0], YDigit, P[YInx], 0);
|
|
Carry := Int32MultPrim(X[1], YDigit, P[YInx + 1], Carry);
|
|
Carry := Int32MultPrim(X[2], YDigit, P[YInx + 2], Carry);
|
|
P[YInx + 4] := Int32MultPrim(X[3], YDigit, P[YInx + 3], Carry);
|
|
end;
|
|
end;
|
|
|
|
{the product has 32 decimal places, so divide by 10^8 twice to get
|
|
the answer to the 16 decimal places we need}
|
|
Int256Div10E8(P, Carry);
|
|
Int256Div10E8(P, Carry);
|
|
|
|
{note: if Carry <> 0 then we're losing precision}
|
|
|
|
{check for multiplication overflow}
|
|
if (P[3] < 0) or
|
|
(P[4] <> 0) or (P[5] <> 0) or (P[6] <> 0) or (P[7] <> 0) then
|
|
raise EStDecMathError.Create(stscDecMathMultOverflowS);
|
|
|
|
{return the value in X, remembering to set the sign}
|
|
X[0] := P[0];
|
|
X[1] := P[1];
|
|
X[2] := P[2];
|
|
X[3] := P[3];
|
|
|
|
(*
|
|
{round if necessary}
|
|
if (Carry >= 500000000) then
|
|
Int128AddInt(X, 1);
|
|
*)
|
|
|
|
{set the sign}
|
|
if (XIsNeg xor YIsNeg) then
|
|
Int128ChgSign(X);
|
|
end;
|
|
{--------}
|
|
procedure Int128TimesInt(var X : TStInt128; aValue : integer);
|
|
{Note: this routine assumes X is positive}
|
|
asm
|
|
push ebx // save ebx
|
|
push ebp // save ebp
|
|
|
|
mov ecx, edx // we're multiplying by aValue
|
|
mov ebx, eax // ebx points to X
|
|
|
|
mov eax, [ebx] // get the first 32-bit digit
|
|
mul ecx // multiply it by 10 to give answer in edx:eax
|
|
mov [ebx], eax // save first digit of result
|
|
mov ebp, edx // save overflow
|
|
|
|
mov eax, [ebx+4] // get the second 32-bit digit
|
|
mul ecx // multiply it by 10 to give answer in edx:eax
|
|
add eax, ebp // add the overflow from the first digit
|
|
adc edx, 0
|
|
mov [ebx+4], eax // save second digit of result
|
|
mov ebp, edx // save overflow
|
|
|
|
mov eax, [ebx+8] // get the third 32-bit digit
|
|
mul ecx // multiply it by 10 to give answer in edx:eax
|
|
add eax, ebp // add the overflow from the second digit
|
|
adc edx, 0
|
|
mov [ebx+8], eax // save second digit of result
|
|
mov ebp, edx // save overflow
|
|
|
|
mov eax, [ebx+12] // get the third 32-bit digit
|
|
mul ecx // multiply it by 10 to give answer in edx:eax
|
|
add eax, ebp // add the overflow from the second digit
|
|
mov [ebx+12], eax // save third digit of result
|
|
|
|
pop ebp // restore ebp
|
|
pop ebx // restore ebx
|
|
end;
|
|
{--------}
|
|
procedure Int128Round(var X : TStInt128;
|
|
aRound : TStRoundMethod;
|
|
aDecPl : integer);
|
|
var
|
|
Rem : integer;
|
|
HadRem : boolean;
|
|
AddOne : boolean;
|
|
Expnt : integer;
|
|
NeedInt : boolean;
|
|
begin
|
|
{Assumptions: X is positive, 0 <= aDecPl <= MaxDecPl
|
|
--the caller *must* ensure these}
|
|
|
|
{if the number of decimal places is -1, it's a special signal to
|
|
perform the rounding to an integer, but not to multiply the result
|
|
by 10^16 at the end; the caller is AsInt, in other words}
|
|
if (aDecPl >= 0) then
|
|
NeedInt := false
|
|
else begin
|
|
NeedInt := true;
|
|
aDecPl := 0;
|
|
end;
|
|
|
|
{if we're asked to round to the precision of the type, there's
|
|
nothing to do}
|
|
if (aDecPl = MaxDecPl) then
|
|
Exit;
|
|
|
|
{perform the required rounding}
|
|
AddOne := false; // keep the compiler happy
|
|
case aRound of
|
|
rmNormal :
|
|
begin
|
|
{to do normal rounding: divide by the required power of ten,
|
|
if the most significant digit of the remainder was 5 or more,
|
|
we'll add one to the result}
|
|
Expnt := MaxDecPl - aDecPl - 1;
|
|
if (Expnt > 0) then begin
|
|
if (Expnt > 8) then begin
|
|
Int128DivInt(X, PowerOf10[8]);
|
|
dec(Expnt, 8);
|
|
end;
|
|
Int128DivInt(X, PowerOf10[Expnt]);
|
|
end;
|
|
AddOne := Int128DivInt(X, 10) >= 5;
|
|
end;
|
|
rmTrunc :
|
|
begin
|
|
{to truncate: just divide by the required power of ten}
|
|
Expnt := MaxDecPl - aDecPl;
|
|
if (Expnt > 8) then begin
|
|
Int128DivInt(X, PowerOf10[8]);
|
|
dec(Expnt, 8);
|
|
end;
|
|
Int128DivInt(X, PowerOf10[Expnt]);
|
|
AddOne := false;
|
|
end;
|
|
rmBankers :
|
|
begin
|
|
{to do bankers rounding:
|
|
- divide by the required power of ten, checking to see if
|
|
there's a non-zero remainder
|
|
- if the most significant digit of the remainder was greater
|
|
than 5, we'll add one to the result
|
|
- if the most significant digit of the remainder was 5 and
|
|
there was at least one other digit in the remainder, we'll
|
|
add one to the result
|
|
- if the most significant digit of the remainder was 5 and
|
|
there were no other digits in the remainder, determine if
|
|
the result is odd; if it is, we'll add one to the result}
|
|
HadRem := false;
|
|
if ((MaxDecPl - aDecPl) > 1) then begin
|
|
Expnt := MaxDecPl - aDecPl - 1;
|
|
if (Expnt > 8) then begin
|
|
if (Int128DivInt(X, PowerOf10[8]) <> 0) then
|
|
HadRem := true;
|
|
dec(Expnt, 8);
|
|
end;
|
|
if (Int128DivInt(X, PowerOf10[Expnt]) <> 0) then
|
|
HadRem := true;
|
|
end;
|
|
Rem := Int128DivInt(X, 10);
|
|
AddOne := (Rem > 5) or
|
|
((Rem = 5) and HadRem) or
|
|
((Rem = 5) and Odd(X[0]));
|
|
end;
|
|
rmUp :
|
|
begin
|
|
{to always round up: divide by the required power of ten,
|
|
if there was a remainder, we'll add one to the result}
|
|
AddOne := false;
|
|
Expnt := MaxDecPl - aDecPl;
|
|
if (Expnt > 8) then begin
|
|
if (Int128DivInt(X, PowerOf10[8]) <> 0) then
|
|
AddOne := true;
|
|
dec(Expnt, 8);
|
|
end;
|
|
if (Int128DivInt(X, PowerOf10[Expnt]) <> 0) then
|
|
AddOne := true;
|
|
end;
|
|
end;{case}
|
|
|
|
{add one to the result, if required}
|
|
if AddOne then
|
|
Int128AddInt(X, 1);
|
|
|
|
{finally, multiply by the required power of ten}
|
|
if not NeedInt then begin
|
|
Expnt := MaxDecPl - aDecPl;
|
|
if (Expnt > 8) then begin
|
|
Int128TimesInt(X, PowerOf10[8]);
|
|
dec(Expnt, 8);
|
|
end;
|
|
Int128TimesInt(X, PowerOf10[Expnt]);
|
|
end;
|
|
end;
|
|
{====================================================================}
|
|
|
|
|
|
{====================================================================}
|
|
constructor TStDecimal.Create;
|
|
begin
|
|
{create the ancestor}
|
|
inherited Create;
|
|
{note: the internal number will be automatically zero}
|
|
end;
|
|
{--------}
|
|
destructor TStDecimal.Destroy;
|
|
begin
|
|
{free the ancestor}
|
|
inherited Destroy;
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.Abs;
|
|
begin
|
|
if (FInt[3] < 0) then
|
|
Int128ChgSign(FInt);
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.Add(X : TStDecimal);
|
|
begin
|
|
if (X <> nil) then
|
|
Int128Add(FInt, X.FInt);
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.AddOne;
|
|
var
|
|
One : TStInt128;
|
|
begin
|
|
One[0] := Int128One_0;
|
|
One[1] := Int128One_1;
|
|
One[2] := 0;
|
|
One[3] := 0;
|
|
Int128Add(FInt, One);
|
|
end;
|
|
{--------}
|
|
function TStDecimal.AsFloat : double;
|
|
begin
|
|
Result := StrToFloat(AsString);
|
|
end;
|
|
{--------}
|
|
function TStDecimal.AsInt(aRound : TStRoundMethod) : integer;
|
|
var
|
|
X : TStInt128;
|
|
IsNeg : boolean;
|
|
begin
|
|
{get the current value locally}
|
|
X[0] := FInt[0];
|
|
X[1] := FInt[1];
|
|
X[2] := FInt[2];
|
|
X[3] := FInt[3];
|
|
|
|
{force it to be positive}
|
|
IsNeg := X[3] < 0;
|
|
if IsNeg then
|
|
Int128ChgSign(X);
|
|
|
|
{round it to an integer}
|
|
Int128Round(X, aRound, -1);
|
|
|
|
{check for errors (the least significant digit cannot be negative,
|
|
and all the others must be zero)}
|
|
if (X[0] < 0) or (X[1] <> 0) or (X[2] <> 0) or (X[3] <> 0) then
|
|
raise EStDecMathError.Create(stscDecMathAsIntOverflowS);
|
|
|
|
{return the result}
|
|
if IsNeg then
|
|
Result := -X[0]
|
|
else
|
|
Result := X[0];
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.Assign(X : TStDecimal);
|
|
begin
|
|
if (X = nil) then
|
|
SetToZero
|
|
else begin
|
|
FInt[0] := X.FInt[0];
|
|
FInt[1] := X.FInt[1];
|
|
FInt[2] := X.FInt[2];
|
|
FInt[3] := X.FInt[3];
|
|
end;
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.AssignFromFloat(aValue : double);
|
|
begin
|
|
AsString := Format('%38.16f', [aValue]);
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.AssignFromInt(aValue : integer);
|
|
begin
|
|
FInt[0] := System.Abs(aValue);
|
|
FInt[1] := 0;
|
|
FInt[2] := 0;
|
|
FInt[3] := 0;
|
|
Int128TimesInt(FInt, PowerOf10[8]);
|
|
Int128TimesInt(FInt, PowerOf10[8]);
|
|
if (aValue < 0) then
|
|
Int128ChgSign(FInt);
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.ChangeSign;
|
|
begin
|
|
Int128ChgSign(FInt);
|
|
end;
|
|
{--------}
|
|
function TStDecimal.Compare(X : TStDecimal) : integer;
|
|
begin
|
|
Compare := Int128Compare(FInt, X.FInt);
|
|
end;
|
|
{--------}
|
|
function TStDecimal.dcGetAsStr : AnsiString;
|
|
var
|
|
X : TStInt128;
|
|
i : integer;
|
|
Rem : integer;
|
|
IsNeg : boolean;
|
|
ChStack: array [0..47] of AnsiChar;
|
|
// this is ample for 38 digits + punctuation
|
|
ChSP : integer;
|
|
begin
|
|
{initialize the stack}
|
|
ChSP := 0;
|
|
|
|
{since we're going to trash the value, store it locally}
|
|
X[0] := FInt[0];
|
|
X[1] := FInt[1];
|
|
X[2] := FInt[2];
|
|
X[3] := FInt[3];
|
|
|
|
{make sure it's positive}
|
|
IsNeg := X[3] < 0;
|
|
if IsNeg then
|
|
Int128ChgSign(X);
|
|
|
|
{push the least significant digits (those that will appear after the
|
|
radix point)}
|
|
for i := 1 to MaxDecPl do begin
|
|
Rem := Int128DivInt(X, 10);
|
|
ChStack[ChSP] := AnsiChar(Rem + ord('0'));
|
|
inc(ChSP);
|
|
end;
|
|
|
|
{push the radix point}
|
|
ChStack[ChSP] := AnsiChar({$IFDEF DELPHIXE2}FormatSettings.{$ENDIF}DecimalSeparator);
|
|
inc(ChSP);
|
|
|
|
{repeat until the local value is zero}
|
|
repeat
|
|
Rem := Int128DivInt(X, 10);
|
|
ChStack[ChSP] := AnsiChar(Rem + ord('0'));
|
|
inc(ChSP);
|
|
until (X[0] = 0) and (X[1] = 0) and (X[2] = 0) and (X[3] = 0);
|
|
|
|
{if the value was negative, push a minus sign}
|
|
if IsNeg then begin
|
|
ChStack[ChSP] := '-';
|
|
inc(ChSP);
|
|
end;
|
|
|
|
{construct the result value by popping off characters}
|
|
SetLength(Result, ChSP);
|
|
i := 1;
|
|
while (ChSP <> 0) do begin
|
|
dec(ChSP);
|
|
Result[i] := ChStack[ChSP];
|
|
inc(i);
|
|
end;
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.dcSetFromStr(const aValue : AnsiString); {!!.02}
|
|
var
|
|
State : (ScanStart, ScanSign, ScanRadix, ScanBefore,
|
|
ScanAfter, ScanEnd, GotError);
|
|
i : integer;
|
|
Ch : AnsiChar;
|
|
IsNeg : boolean;
|
|
DecPlCount : integer;
|
|
begin
|
|
{Note: this implements the following DFA:
|
|
|
|
ScanStart --space--> ScanStart
|
|
ScanStart --plus---> ScanSign
|
|
ScanStart --minus--> ScanSign
|
|
ScanStart --digit--> ScanBefore
|
|
ScanStart --radix--> ScanRadix
|
|
|
|
ScanSign --radix--> ScanRadix
|
|
ScanSign --digit--> ScanBefore
|
|
|
|
ScanRadix --digit--> ScanAfter
|
|
|
|
ScanBefore --radix--> ScanAfter
|
|
ScanBefore --digit--> ScanBefore
|
|
ScanBefore --space--> ScanEnd
|
|
|
|
ScanAfter --digit--> ScanAfter
|
|
ScanAfter --space--> ScanEnd
|
|
|
|
ScanEnd --space--> ScanEnd
|
|
|
|
The terminating states are ScanBefore, ScanAfter and ScanEnd; in
|
|
other words, a valid numeric string cannot end in a radix point.
|
|
}
|
|
|
|
{initialize}
|
|
SetToZero;
|
|
DecPlCount := 0;
|
|
IsNeg := false;
|
|
State := ScanStart;
|
|
|
|
{read through the input string}
|
|
for i := 1 to length(aValue) do begin
|
|
|
|
{get the current character}
|
|
Ch := aValue[i];
|
|
|
|
case State of
|
|
ScanStart :
|
|
begin
|
|
if ('0' <= Ch) and (Ch <= '9') then begin
|
|
FInt[0] := ord(Ch) - ord('0');
|
|
State := ScanBefore;
|
|
end
|
|
else if (Ch = '+') then begin
|
|
State := ScanSign;
|
|
end
|
|
else if (Ch = '-') then begin
|
|
IsNeg := true;
|
|
State := ScanSign;
|
|
end
|
|
else if (Ch = AnsiChar({$IFDEF DELPHIXE2}FormatSettings.{$ENDIF}DecimalSeparator)) then begin
|
|
State := ScanRadix;
|
|
end
|
|
else if (Ch <> ' ') then
|
|
State := GotError;
|
|
end;
|
|
ScanSign :
|
|
begin
|
|
if ('0' <= Ch) and (Ch <= '9') then begin
|
|
FInt[0] := ord(Ch) - ord('0');
|
|
State := ScanBefore;
|
|
end
|
|
else if (Ch = AnsiChar({$IFDEF DELPHIXE2}FormatSettings.{$ENDIF}DecimalSeparator)) then begin
|
|
State := ScanRadix;
|
|
end
|
|
else
|
|
State := GotError;
|
|
end;
|
|
ScanRadix :
|
|
begin
|
|
if ('0' <= Ch) and (Ch <= '9') then begin
|
|
inc(DecPlCount);
|
|
Int128TimesInt(FInt, 10);
|
|
Int128AddInt(FInt, ord(Ch) - ord('0'));
|
|
State := ScanAfter;
|
|
end
|
|
else
|
|
State := GotError;
|
|
end;
|
|
ScanBefore :
|
|
begin
|
|
if ('0' <= Ch) and (Ch <= '9') then begin
|
|
Int128TimesInt(FInt, 10);
|
|
Int128AddInt(FInt, ord(Ch) - ord('0'));
|
|
end
|
|
else if (Ch = AnsiChar({$IFDEF DELPHIXE2}FormatSettings.{$ENDIF}DecimalSeparator)) then begin
|
|
State := ScanAfter;
|
|
end
|
|
else if (Ch = ' ') then
|
|
State := ScanEnd
|
|
else
|
|
State := GotError;
|
|
end;
|
|
ScanAfter :
|
|
begin
|
|
if ('0' <= Ch) and (Ch <= '9') then begin
|
|
inc(DecPlCount);
|
|
if (DecPlCount <= MaxDecPl) then begin
|
|
Int128TimesInt(FInt, 10);
|
|
Int128AddInt(FInt, ord(Ch) - ord('0'));
|
|
end;
|
|
end
|
|
else if (Ch = ' ') then
|
|
State := ScanEnd
|
|
else
|
|
State := GotError;
|
|
end;
|
|
ScanEnd :
|
|
begin
|
|
if (Ch <> ' ') then
|
|
State := GotError;
|
|
end;
|
|
GotError :
|
|
begin
|
|
Break;
|
|
end;
|
|
end;
|
|
end;
|
|
|
|
if (State <> ScanBefore) and
|
|
(State <> ScanAfter) and
|
|
(State <> ScanEnd) then
|
|
raise EStDecMathError.Create(stscDecMathConversionS);
|
|
|
|
{make sure we have the correct number of decimal places}
|
|
if (MaxDecPl > DecPlCount) then begin
|
|
DecPlCount := MaxDecPl - DecPlCount;
|
|
if (DecPlCount > 8) then begin
|
|
Int128TimesInt(FInt, Powerof10[8]);
|
|
dec(DecPlCount, 8);
|
|
end;
|
|
Int128TimesInt(FInt, Powerof10[DecPlCount]);
|
|
end;
|
|
|
|
{force negative, if required}
|
|
if IsNeg then
|
|
Int128ChgSign(FInt);
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.Divide(X : TStDecimal);
|
|
var
|
|
TempX : TStInt128;
|
|
IsNeg : boolean;
|
|
XIsNeg : boolean;
|
|
begin
|
|
{easy case: X is nil or zero}
|
|
if (X = nil) or X.IsZero then
|
|
raise EStDecMathError.Create(stscDecMathDivByZeroS);
|
|
|
|
{easy case: Self is zero}
|
|
if IsZero then
|
|
Exit;
|
|
|
|
{we might have to change X, so make it local}
|
|
TempX[0] := X.FInt[0];
|
|
TempX[1] := X.FInt[1];
|
|
TempX[2] := X.FInt[2];
|
|
TempX[3] := X.FInt[3];
|
|
|
|
{force the divisor and dividend positive}
|
|
IsNeg := FInt[3] < 0;
|
|
if IsNeg then
|
|
Int128ChgSign(FInt);
|
|
XIsNeg := TempX[3] < 0;
|
|
if XIsNeg then
|
|
Int128ChgSign(TempX);
|
|
|
|
{easy case: X is 1.0: set the correct sign}
|
|
if (TempX[0] = Int128One_0) and (TempX[1] = Int128One_1) and
|
|
(TempX[2] = 0) and (TempX[3] = 0) then begin
|
|
if (IsNeg xor XIsNeg) then
|
|
Int128ChgSign(FInt);
|
|
Exit;
|
|
end;
|
|
|
|
{easy case: compare the dividend and divisor: if they're equal,
|
|
set ourselves to 1.0 with the correct sign}
|
|
if (Int128Compare(FInt, TempX) = 0) then begin
|
|
FInt[0] := Int128One_0;
|
|
FInt[1] := Int128One_1;
|
|
FInt[2] := 0;
|
|
FInt[3] := 0;
|
|
if (IsNeg xor XIsNeg) then
|
|
Int128ChgSign(FInt);
|
|
Exit;
|
|
end;
|
|
|
|
{no more easy cases: just do the division}
|
|
Int128Divide(FInt, TempX);
|
|
|
|
{set the sign}
|
|
if (IsNeg xor XIsNeg) then
|
|
Int128ChgSign(FInt);
|
|
end;
|
|
{--------}
|
|
function TStDecimal.IsNegative : boolean;
|
|
begin
|
|
{if the most significant longint is negative, so is the value}
|
|
Result := FInt[3] < 0;
|
|
end;
|
|
{--------}
|
|
function TStDecimal.IsOne : boolean;
|
|
begin
|
|
Result := (FInt[0] = Int128One_0) and (FInt[1] = Int128One_1) and
|
|
(FInt[2] = 0) and (FInt[3] = 0);
|
|
end;
|
|
{--------}
|
|
function TStDecimal.IsPositive : boolean;
|
|
begin
|
|
{if the most significant longint is positive, so is the value; if it
|
|
is zero, one of the other longints must be non-zero for the value
|
|
to be positive}
|
|
Result := (FInt[3] > 0) or
|
|
((FInt[3] = 0) and
|
|
((FInt[2] <> 0) or (FInt[1] <> 0) or (FInt[0] <> 0)));
|
|
end;
|
|
{--------}
|
|
function TStDecimal.IsZero : boolean;
|
|
begin
|
|
Result := (FInt[0] = 0) and (FInt[1] = 0) and
|
|
(FInt[2] = 0) and (FInt[3] = 0);
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.Multiply(X : TStDecimal);
|
|
begin
|
|
if (X = nil) then
|
|
SetToZero
|
|
else
|
|
Int128Multiply(FInt, X.FInt);
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.RaiseToPower(N : integer);
|
|
var
|
|
Accum : TStInt128;
|
|
Mask : longint;
|
|
IsNeg : boolean;
|
|
begin
|
|
{take care of some easy cases}
|
|
if (N < 0) then
|
|
raise EStDecMathError.Create(stscDecMathNegExpS);
|
|
if (N = 0) then begin
|
|
SetToOne;
|
|
Exit;
|
|
end;
|
|
if (N = 1) then
|
|
Exit;
|
|
|
|
{force the value positive}
|
|
IsNeg := FInt[3] < 0;
|
|
if IsNeg then
|
|
Int128ChgSign(FInt);
|
|
|
|
{initialize the accumulator to 1.0}
|
|
Accum[0] := Int128One_0;
|
|
Accum[1] := Int128One_1;
|
|
Accum[2] := 0;
|
|
Accum[3] := 0;
|
|
|
|
{set the bit mask}
|
|
Mask := longint($80000000);
|
|
|
|
{find the first set bit in the exponent}
|
|
while ((N and Mask) = 0) do
|
|
Mask := Mask shr 1;
|
|
|
|
{calculate the power}
|
|
while (Mask <> 0) do begin
|
|
Int128Multiply(Accum, Accum);
|
|
if ((N and Mask) <> 0) then
|
|
Int128Multiply(Accum, FInt);
|
|
Mask := Mask shr 1;
|
|
end;
|
|
|
|
{save the calculated value}
|
|
FInt[0] := Accum[0];
|
|
FInt[1] := Accum[1];
|
|
FInt[2] := Accum[2];
|
|
FInt[3] := Accum[3];
|
|
|
|
{force the value negative if required}
|
|
if IsNeg and Odd(N) then
|
|
Int128ChgSign(FInt);
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.Round(aRound : TStRoundMethod; aDecPl : integer);
|
|
var
|
|
IsNeg : boolean;
|
|
begin
|
|
{check decimal places parameter to be in range}
|
|
if not ((0 <= aDecPl) and (aDecPl <= MaxDecPl)) then
|
|
raise EStDecMathError.Create(stscDecMathRoundPlacesS);
|
|
|
|
{force the value positive}
|
|
IsNeg := FInt[3] < 0;
|
|
if IsNeg then
|
|
Int128ChgSign(FInt);
|
|
|
|
{perform the rounding}
|
|
Int128Round(FInt, aRound, aDecPl);
|
|
|
|
{force the value negative if it was negative}
|
|
if IsNeg then
|
|
Int128ChgSign(FInt);
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.SetToOne;
|
|
begin
|
|
FInt[0] := Int128One_0;
|
|
FInt[1] := Int128One_1;
|
|
FInt[2] := 0;
|
|
FInt[3] := 0;
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.SetToZero;
|
|
begin
|
|
FInt[0] := 0;
|
|
FInt[1] := 0;
|
|
FInt[2] := 0;
|
|
FInt[3] := 0;
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.Subtract(X : TStDecimal);
|
|
var
|
|
MinusX : TStInt128;
|
|
begin
|
|
if (X <> nil) then begin
|
|
MinusX[0] := X.FInt[0];
|
|
MinusX[1] := X.FInt[1];
|
|
MinusX[2] := X.FInt[2];
|
|
MinusX[3] := X.FInt[3];
|
|
Int128ChgSign(MinusX);
|
|
Int128Add(Fint, MinusX);
|
|
end;
|
|
end;
|
|
{--------}
|
|
procedure TStDecimal.SubtractOne;
|
|
var
|
|
MinusOne : TStInt128;
|
|
begin
|
|
MinusOne[0] := Int128One_0;
|
|
MinusOne[1] := Int128One_1;
|
|
MinusOne[2] := 0;
|
|
MinusOne[3] := 0;
|
|
Int128ChgSign(MinusOne);
|
|
Int128Add(FInt, MinusOne);
|
|
end;
|
|
{====================================================================}
|
|
|
|
end.
|