4
0
mirror of git://sourceware.org/git/newlib-cygwin.git synced 2025-01-23 07:27:21 +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

119 lines
3.8 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_hypot.c in Newlib. */
#include "amdgcnmach.h"
v64sf v64sf_sqrtf_aux (v64sf, v64si);
DEF_VS_MATH_FUNC (v64sf, hypotf, v64sf x, v64sf y)
{
FUNCTION_INIT (v64sf);
v64sf a = x;
v64sf b = y;
v64si ha;
GET_FLOAT_WORD (ha, x, NO_COND);
ha &= 0x7fffffffL;
v64si hb;
GET_FLOAT_WORD (hb, y, NO_COND);
hb &= 0x7fffffffL;
VECTOR_IF (hb > ha, cond)
v64si j = ha;
VECTOR_COND_MOVE (ha, hb, cond);
VECTOR_COND_MOVE (hb, j, cond);
VECTOR_ENDIF
SET_FLOAT_WORD (a, ha, NO_COND); /* a <- |a| */
SET_FLOAT_WORD (b, hb, NO_COND); /* b <- |b| */
VECTOR_IF((ha - hb) > 0xf000000L, cond) // x/y > 2**30 */
VECTOR_RETURN (a + b, cond);
VECTOR_ENDIF
v64si k = VECTOR_INIT (0);
VECTOR_IF (ha > 0x58800000L, cond) /* a>2**50 */
VECTOR_IF2 (ha >= 0x7f800000L, cond2, cond) /* Inf or NaN */
v64sf w = a + b; // for sNaN */
VECTOR_COND_MOVE (w, a, cond2 & (ha == 0x7f800000));
VECTOR_COND_MOVE (w, b, cond2 & (hb == 0x7f800000));
VECTOR_RETURN (w, cond);
VECTOR_ENDIF
/* scale a and b by 2**-60 */
VECTOR_COND_MOVE (ha, ha - 0x5d800000, cond);
VECTOR_COND_MOVE (hb, hb - 0x5d800000, cond);
VECTOR_COND_MOVE (k, k + 60, cond);
SET_FLOAT_WORD (a, ha, cond);
SET_FLOAT_WORD (b, hb, cond);
VECTOR_ENDIF
VECTOR_IF (hb < 0x26800000, cond) /* b < 2**-50 */
VECTOR_IF2 (hb <= 0x007fffff, cond2, cond) /* subnormal b or 0 */
VECTOR_RETURN (a, cond2 & (hb == 0));
/* t1=2^126 */
v64sf t1;
SET_FLOAT_WORD (t1, VECTOR_INIT (0x3f000000), cond2);
VECTOR_COND_MOVE (b, b * t1, cond2);
VECTOR_COND_MOVE (a, a * t1, cond2);
VECTOR_COND_MOVE (k, k - 126, cond2);
VECTOR_ELSE2 (cond2, cond) /* scale a and b by 2^60 */
VECTOR_COND_MOVE (ha, ha + 0x5d800000, cond2); /* a *= 2^60 */
VECTOR_COND_MOVE (hb, hb + 0x5d800000, cond2); /* b *= 2^60 */
VECTOR_COND_MOVE (k, k - 60, cond2);
SET_FLOAT_WORD (a, ha, cond2);
SET_FLOAT_WORD (b, hb, cond2);
VECTOR_ENDIF
VECTOR_ENDIF
/* medium size a and b */
v64sf w = a - b;
VECTOR_IF (w > b, cond)
v64sf t1;
SET_FLOAT_WORD (t1, ha & 0xfffff000, cond);
v64sf t2 = a - t1;
VECTOR_COND_MOVE (w, v64sf_sqrtf_aux (t1*t1 - (b*(-b) - t2 * (a + t1)), __mask), cond);
VECTOR_ELSE (cond)
VECTOR_COND_MOVE (a, a+a, cond);
v64sf y1;
SET_FLOAT_WORD (y1, hb & 0xfffff000, cond);
v64sf y2 = b - y1;
v64sf t1;
SET_FLOAT_WORD (t1, ha + 0x00800000, cond);
v64sf t2 = a - t1;
VECTOR_COND_MOVE (w, v64sf_sqrtf_aux (t1*y1 - (w*(-w) - (t1*y2 + t2*b)), __mask), cond);
VECTOR_ENDIF
VECTOR_IF (k != 0, cond)
v64sf t1;
SET_FLOAT_WORD (t1, 0x3f800000 + (k << 23), cond);
VECTOR_RETURN (t1 * w, cond);
VECTOR_ELSE (cond)
VECTOR_RETURN (w, cond);
VECTOR_ENDIF
FUNCTION_RETURN;
}
DEF_VARIANTS2 (hypotf, sf, sf)