93 lines
3.1 KiB
C
93 lines
3.1 KiB
C
/*
|
|
* Copyright 2023 Siemens
|
|
*
|
|
* The authors hereby grant permission to use, copy, modify, distribute,
|
|
* and license this software and its documentation for any purpose, provided
|
|
* that existing copyright notices are retained in all copies and that this
|
|
* notice is included verbatim in any distributions. No written agreement,
|
|
* license, or royalty fee is required for any of the authorized uses.
|
|
* Modifications to this software may be copyrighted by their authors
|
|
* and need not follow the licensing terms described here, provided that
|
|
* the new terms are clearly indicated on the first page of each file where
|
|
* they apply.
|
|
*/
|
|
|
|
/*
|
|
* Copyright (c) 1994-2009 Red Hat, Inc. All rights reserved.
|
|
*
|
|
* This copyrighted material is made available to anyone wishing to use,
|
|
* modify, copy, or redistribute it subject to the terms and conditions
|
|
* of the BSD License. This program is distributed in the hope that
|
|
* it will be useful, but WITHOUT ANY WARRANTY expressed or implied,
|
|
* including the implied warranties of MERCHANTABILITY or FITNESS FOR
|
|
* A PARTICULAR PURPOSE. A copy of this license is available at
|
|
* http://www.opensource.org/licenses. Any Red Hat trademarks that are
|
|
* incorporated in the source code or documentation are not subject to
|
|
* the BSD License and may only be used or replicated with the express
|
|
* permission of Red Hat, Inc.
|
|
*/
|
|
|
|
/*****************************************************************
|
|
* The following routines are coded directly from the algorithms
|
|
* and coefficients given in "Software Manual for the Elementary
|
|
* Functions" by William J. Cody, Jr. and William Waite, Prentice
|
|
* Hall, 1980.
|
|
*****************************************************************/
|
|
|
|
/* Based on newlib/libm/mathfp/s_tanh.c in Newlib. */
|
|
|
|
#include "amdgcnmach.h"
|
|
|
|
v64df v64df_exp_aux (v64df, v64di);
|
|
|
|
static const double LN3_OVER2 = 0.54930614433405484570;
|
|
static const double p[] = { -0.16134119023996228053e+4,
|
|
-0.99225929672236083313e+2,
|
|
-0.96437492777225469787 };
|
|
static const double q[] = { 0.48402357071988688686e+4,
|
|
0.22337720718962312926e+4,
|
|
0.11274474380534949335e+3 };
|
|
|
|
#if defined (__has_builtin) && __has_builtin (__builtin_gcn_fabsv)
|
|
|
|
DEF_VD_MATH_FUNC (v64df, tanh, v64df x)
|
|
{
|
|
FUNCTION_INIT (v64df);
|
|
|
|
v64df f = __builtin_gcn_fabsv (x);
|
|
v64df res;
|
|
|
|
/* Check if the input is too big. */
|
|
VECTOR_IF (f > BIGX, cond)
|
|
VECTOR_COND_MOVE (res, VECTOR_INIT (1.0), cond);
|
|
|
|
VECTOR_ELSEIF (f > LN3_OVER2, cond)
|
|
VECTOR_COND_MOVE (res, 1.0 - 2.0 / (v64df_exp_aux (2 * f, __mask) + 1.0),
|
|
cond);
|
|
|
|
/* Check if the input is too small. */
|
|
VECTOR_ELSEIF (f < z_rooteps, cond)
|
|
VECTOR_COND_MOVE (res, f, cond);
|
|
|
|
/* Calculate the Taylor series. */
|
|
VECTOR_ELSE (cond)
|
|
v64df g = f * f;
|
|
|
|
v64df P = (p[2] * g + p[1]) * g + p[0];
|
|
v64df Q = ((g + q[2]) * g + q[1]) * g + q[0];
|
|
v64df R = g * (P / Q);
|
|
|
|
VECTOR_COND_MOVE (res, f + f * R, cond);
|
|
VECTOR_ENDIF
|
|
|
|
VECTOR_COND_MOVE (res, -res, x < 0.0);
|
|
|
|
VECTOR_RETURN (res, NO_COND);
|
|
|
|
FUNCTION_RETURN;
|
|
}
|
|
|
|
DEF_VARIANTS (tanh, df, df)
|
|
|
|
#endif
|