4
0
mirror of git://sourceware.org/git/newlib-cygwin.git synced 2025-01-23 23:47:22 +08:00
Kwok Cheung Yeung e18743072b amdgcn: Add vectorized math routines
This implements a set of vectorized math routines to be used by the
compiler auto-vectorizer.  Versions for vectors with 2 lanes up to
64 lanes (in powers of 2) are provided.

These routines are based on the scalar versions of the math routines in
libm/common, libm/math and libm/mathfp.  They make extensive use of the GCC
C vector extensions and GCN-specific builtins in GCC.
2023-01-18 13:22:58 -05:00

81 lines
2.4 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) 1993 by Sun Microsystems, Inc. All rights reserved.
*
* Developed at SunPro, a Sun Microsystems, Inc. business.
* Permission to use, copy, modify, and distribute this
* software is freely granted, provided that this notice
* is preserved.
* ====================================================
*
*/
/* Based on newlib/libm/mathfp/ef_acosh.c in Newlib. */
#include "amdgcnmach.h"
v64sf v64sf_logf_aux (v64sf, v64si);
v64sf v64sf_log1pf_aux (v64sf, v64si);
v64sf v64sf_sqrtf_aux (v64sf, v64si);
DEF_VS_MATH_FUNC (v64sf, acoshf, v64sf x)
{
static const float one = 1.0;
static const float ln2 = 6.9314718246e-01; /* 0x3f317218 */
FUNCTION_INIT (v64sf);
v64si hx;
GET_FLOAT_WORD (hx, x, NO_COND);
VECTOR_IF (hx < 0x3f800000, cond) // x < 1 */
VECTOR_RETURN ((x-x) / (x-x), cond);
VECTOR_ENDIF
VECTOR_IF (hx >=0x4d800000, cond) // x > 2**28 */
VECTOR_IF2 (hx >=0x7f800000, cond2, cond) // x is inf of NaN */
VECTOR_RETURN (x+x, cond2);
VECTOR_ELSE (cond2)
/* acosh(huge)=log(2x) */
VECTOR_RETURN (v64sf_logf_aux (x, __mask) + ln2, cond2);
VECTOR_ENDIF
VECTOR_ENDIF
VECTOR_IF (hx == 0x3f800000, cond)
/* acosh(1) = 0 */
VECTOR_RETURN (VECTOR_INIT (0.0f), cond);
VECTOR_ENDIF
VECTOR_IF (hx > 0x40000000, cond) /* 2**28 > x > 2 */
{
v64sf t = x * x;
VECTOR_RETURN (v64sf_logf_aux (2.0f*x - 1.0f /
(x + v64sf_sqrtf_aux (t - 1.0f, __mask)),
__mask),
cond);
}
VECTOR_ELSE (cond) /* 1<x<2 */
{
v64sf t = x - 1.0f;
VECTOR_RETURN (v64sf_log1pf_aux (t + v64sf_sqrtf_aux(2.0*t + t*t, __mask),
__mask),
cond);
}
VECTOR_ENDIF
FUNCTION_RETURN;
}
DEF_VARIANTS (acoshf, sf, sf)