232 lines
5.3 KiB
C++
232 lines
5.3 KiB
C++
// Copyright (c) Microsoft Corporation. All rights reserved.
|
|
// Licensed under the MIT License.
|
|
|
|
//-----------------------------------------------------------------------------
|
|
// Package Title ratpak
|
|
// File transh.c
|
|
// Copyright (C) 1995-96 Microsoft
|
|
// Date 01-16-95
|
|
//
|
|
//
|
|
// Description
|
|
//
|
|
// Contains hyperbolic sin, cos, and tan for rationals.
|
|
//
|
|
//
|
|
//-----------------------------------------------------------------------------
|
|
#include "pch.h"
|
|
#include "ratpak.h"
|
|
|
|
|
|
|
|
bool IsValidForHypFunc(PRAT px, int32_t precision)
|
|
{
|
|
PRAT ptmp = nullptr;
|
|
bool bRet = true;
|
|
|
|
DUPRAT(ptmp,rat_min_exp);
|
|
divrat(&ptmp, rat_ten, precision);
|
|
if ( rat_lt( px, ptmp, precision) )
|
|
{
|
|
bRet = false;
|
|
}
|
|
destroyrat( ptmp );
|
|
return bRet;
|
|
|
|
}
|
|
|
|
//-----------------------------------------------------------------------------
|
|
//
|
|
// FUNCTION: sinhrat, _sinhrat
|
|
//
|
|
// ARGUMENTS: x PRAT representation of number to take the sine hyperbolic
|
|
// of
|
|
// RETURN: sinh of x in PRAT form.
|
|
//
|
|
// EXPLANATION: This uses Taylor series
|
|
//
|
|
// n
|
|
// ___ 2j+1
|
|
// \ ] X
|
|
// \ ---------
|
|
// / (2j+1)!
|
|
// /__]
|
|
// j=0
|
|
// or,
|
|
// n
|
|
// ___ 2
|
|
// \ ] X
|
|
// \ thisterm ; where thisterm = thisterm * ---------
|
|
// / j j+1 j (2j)*(2j+1)
|
|
// /__]
|
|
// j=0
|
|
//
|
|
// thisterm = X ; and stop when thisterm < precision used.
|
|
// 0 n
|
|
//
|
|
// if x is bigger than 1.0 (e^x-e^-x)/2 is used.
|
|
//
|
|
//-----------------------------------------------------------------------------
|
|
|
|
|
|
void _sinhrat( PRAT *px, int32_t precision)
|
|
|
|
{
|
|
if ( !IsValidForHypFunc(*px, precision))
|
|
{
|
|
// Don't attempt exp of anything large or small
|
|
throw( CALC_E_DOMAIN );
|
|
}
|
|
|
|
CREATETAYLOR();
|
|
|
|
DUPRAT(pret,*px);
|
|
DUPRAT(thisterm,pret);
|
|
|
|
DUPNUM(n2,num_one);
|
|
|
|
do {
|
|
NEXTTERM(xx,INC(n2) DIVNUM(n2) INC(n2) DIVNUM(n2), precision);
|
|
} while ( !SMALL_ENOUGH_RAT( thisterm, precision) );
|
|
|
|
DESTROYTAYLOR();
|
|
}
|
|
|
|
void sinhrat( PRAT *px, uint32_t radix, int32_t precision)
|
|
|
|
{
|
|
PRAT tmpx= nullptr;
|
|
|
|
if ( rat_ge( *px, rat_one, precision) )
|
|
{
|
|
DUPRAT(tmpx,*px);
|
|
exprat(px, radix, precision);
|
|
tmpx->pp->sign *= -1;
|
|
exprat(&tmpx, radix, precision);
|
|
subrat( px, tmpx, precision);
|
|
divrat( px, rat_two, precision);
|
|
destroyrat( tmpx );
|
|
}
|
|
else
|
|
{
|
|
_sinhrat( px, precision);
|
|
}
|
|
}
|
|
|
|
//-----------------------------------------------------------------------------
|
|
//
|
|
// FUNCTION: coshrat
|
|
//
|
|
// ARGUMENTS: x PRAT representation of number to take the cosine
|
|
// hyperbolic of
|
|
//
|
|
// RETURN: cosh of x in PRAT form.
|
|
//
|
|
// EXPLANATION: This uses Taylor series
|
|
//
|
|
// n
|
|
// ___ 2j
|
|
// \ ] X
|
|
// \ ---------
|
|
// / (2j)!
|
|
// /__]
|
|
// j=0
|
|
// or,
|
|
// n
|
|
// ___ 2
|
|
// \ ] X
|
|
// \ thisterm ; where thisterm = thisterm * ---------
|
|
// / j j+1 j (2j)*(2j+1)
|
|
// /__]
|
|
// j=0
|
|
//
|
|
// thisterm = 1 ; and stop when thisterm < precision used.
|
|
// 0 n
|
|
//
|
|
// if x is bigger than 1.0 (e^x+e^-x)/2 is used.
|
|
//
|
|
//-----------------------------------------------------------------------------
|
|
|
|
|
|
void _coshrat( PRAT *px, uint32_t radix, int32_t precision)
|
|
|
|
{
|
|
if ( !IsValidForHypFunc(*px, precision))
|
|
{
|
|
// Don't attempt exp of anything large or small
|
|
throw( CALC_E_DOMAIN );
|
|
}
|
|
|
|
CREATETAYLOR();
|
|
|
|
pret->pp=longtonum( 1L, radix);
|
|
pret->pq=longtonum( 1L, radix);
|
|
|
|
DUPRAT(thisterm,pret)
|
|
|
|
n2=longtonum(0L, radix);
|
|
|
|
do {
|
|
NEXTTERM(xx,INC(n2) DIVNUM(n2) INC(n2) DIVNUM(n2), precision);
|
|
} while ( !SMALL_ENOUGH_RAT( thisterm, precision) );
|
|
|
|
DESTROYTAYLOR();
|
|
}
|
|
|
|
void coshrat( PRAT *px, uint32_t radix, int32_t precision)
|
|
|
|
{
|
|
PRAT tmpx= nullptr;
|
|
|
|
(*px)->pp->sign = 1;
|
|
(*px)->pq->sign = 1;
|
|
if ( rat_ge( *px, rat_one, precision) )
|
|
{
|
|
DUPRAT(tmpx,*px);
|
|
exprat(px, radix, precision);
|
|
tmpx->pp->sign *= -1;
|
|
exprat(&tmpx, radix, precision);
|
|
addrat( px, tmpx, precision);
|
|
divrat( px, rat_two, precision);
|
|
destroyrat( tmpx );
|
|
}
|
|
else
|
|
{
|
|
_coshrat( px, radix, precision);
|
|
}
|
|
// Since *px might be epsilon below 1 due to TRIMIT
|
|
// we need this trick here.
|
|
if ( rat_lt(*px, rat_one, precision) )
|
|
{
|
|
DUPRAT(*px,rat_one);
|
|
}
|
|
}
|
|
|
|
//-----------------------------------------------------------------------------
|
|
//
|
|
// FUNCTION: tanhrat
|
|
//
|
|
// ARGUMENTS: x PRAT representation of number to take the tangent
|
|
// hyperbolic of
|
|
//
|
|
// RETURN: tanh of x in PRAT form.
|
|
//
|
|
// EXPLANATION: This uses sinhrat and coshrat
|
|
//
|
|
//-----------------------------------------------------------------------------
|
|
|
|
void tanhrat( PRAT *px, uint32_t radix, int32_t precision)
|
|
|
|
{
|
|
PRAT ptmp= nullptr;
|
|
|
|
DUPRAT(ptmp,*px);
|
|
sinhrat(px, radix, precision);
|
|
coshrat(&ptmp, radix, precision);
|
|
mulnumx(&((*px)->pp),ptmp->pq);
|
|
mulnumx(&((*px)->pq),ptmp->pp);
|
|
|
|
destroyrat(ptmp);
|
|
|
|
}
|