newlib-cygwin/newlib/libm/machine/amdgcn/v64df_fmod.c

200 lines
6.6 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/s_fmod.c in Newlib. */
#include "amdgcnmach.h"
DEF_VD_MATH_FUNC (v64df, fmod, v64df x, v64df y)
{
FUNCTION_INIT(v64df);
v64si hx, hy, hz;
v64usi lx, ly, lz;
EXTRACT_WORDS (hx, lx, x);
EXTRACT_WORDS (hy, ly, y);
v64si sx = hx & 0x80000000; /* sign of x */
hx ^=sx; /* |x| */
hy &= 0x7fffffff; /* |y| */
v64df zeroes = VECTOR_MERGE (VECTOR_INIT (-0.0),
VECTOR_INIT (0.0),
sx != 0);
/* purge off exception values */
VECTOR_IF (((hy | ly) == 0) | (hx >= 0x7ff00000)
| ((hy | ((ly | -ly) >> 31)) > 0x7ff00000), cond) // y=0, or x not finite or y is NaN
VECTOR_RETURN ((x * y) / (x * y), cond);
VECTOR_ENDIF
VECTOR_IF (hx <= hy, cond) // |x|<|y| return x
VECTOR_IF2 ((hx < hy) | (lx < ly), cond2, cond)
VECTOR_RETURN (x, cond);
VECTOR_ENDIF
VECTOR_IF2 (lx == ly, cond2, cond)
VECTOR_RETURN (zeroes, cond2);
VECTOR_ENDIF
VECTOR_ENDIF
/* determine ix = ilogb(x) */
v64si ix;
VECTOR_IF (hx < 0x00100000, cond) // subnormal x
VECTOR_IF2 (hx == 0, cond2, cond)
ix = VECTOR_INIT (-1043);
for (v64si i = __builtin_convertvector (lx, v64si);
!ALL_ZEROES_P (cond2 & (i > 0));
i <<= 1)
VECTOR_COND_MOVE (ix, ix - 1, cond2 & (i > 0));
VECTOR_ELSE2 (cond2, cond)
ix = VECTOR_INIT (-1022);
for (v64si i = __builtin_convertvector (hx << 11, v64si);
!ALL_ZEROES_P (cond2 & (i > 0));
i <<= 1)
VECTOR_COND_MOVE (ix, ix - 1, cond2 & (i > 0));
VECTOR_ENDIF
VECTOR_ELSE (cond)
VECTOR_COND_MOVE (ix, (hx >> 20) - 1023, cond);
VECTOR_ENDIF
/* determine iy = ilogb(y) */
v64si iy;
VECTOR_IF (hy < 0x00100000, cond) // subnormal y
VECTOR_IF2 (hy == 0, cond2, cond)
iy = VECTOR_INIT (-1043);
for (v64si i = __builtin_convertvector (ly, v64si);
!ALL_ZEROES_P (cond2 & (i > 0));
i <<= 1)
VECTOR_COND_MOVE (iy, iy - 1, cond2 & (i > 0));
VECTOR_ELSE2 (cond2, cond)
iy = VECTOR_INIT (-1022);
for (v64si i = __builtin_convertvector (hy << 11, v64si);
!ALL_ZEROES_P (cond2 & (i > 0));
i <<= 1)
VECTOR_COND_MOVE (iy, iy - 1, cond2 & (i > 0));
VECTOR_ENDIF
VECTOR_ELSE (cond)
VECTOR_COND_MOVE (iy, (hy >> 20) - 1023, cond);
VECTOR_ENDIF
/* set up {hx,lx}, {hy,ly} and align y to x */
VECTOR_IF (ix >= -1022, cond)
VECTOR_COND_MOVE (hx, 0x00100000 | (0x000fffff & hx), cond);
VECTOR_ELSE (cond) // subnormal x, shift x to normal
{
v64si n = -1022 - ix;
VECTOR_IF2 (n <= 31, cond2, cond)
VECTOR_COND_MOVE (hx, (hx << n) | (lx >> (32 - n)), cond2);
VECTOR_COND_MOVE (lx, lx << n, cond2);
VECTOR_ELSE2 (cond2, cond)
VECTOR_COND_MOVE (hx, __builtin_convertvector (lx << (n - 32), v64si), cond2);
VECTOR_COND_MOVE (lx, VECTOR_INIT (0U), cond2);
VECTOR_ENDIF
}
VECTOR_ENDIF
VECTOR_IF (iy >= -1022, cond)
VECTOR_COND_MOVE (hy, 0x00100000 | (0x000fffff & hy), cond);
VECTOR_ELSE (cond) // subnormal y, shift y to normal
{
v64si n = -1022 - iy;
VECTOR_IF2 (n <= 31, cond2, cond)
VECTOR_COND_MOVE (hy, (hy << n) | (ly >> (32 - n)), cond2);
VECTOR_COND_MOVE (ly, ly << n, cond2);
VECTOR_ELSE2 (cond2, cond)
VECTOR_COND_MOVE (hy, __builtin_convertvector (ly << (n - 32), v64si), cond2);
VECTOR_COND_MOVE (ly, VECTOR_INIT (0U), cond2);
VECTOR_ENDIF
}
VECTOR_ENDIF
/* fix point fmod */
v64si n = ix - iy;
v64si cond = n != 0;
while (!ALL_ZEROES_P (cond))
{
hz = hx - hy;
lz = lx - ly;
VECTOR_IF2 (lx < ly, cond2, cond)
VECTOR_COND_MOVE (hz, hz - 1, cond2);
VECTOR_ENDIF
VECTOR_IF2 (hz < 0, cond2, cond)
VECTOR_COND_MOVE (hx, hx + hx + (__builtin_convertvector(lx, v64usi) >> 31), cond2);
VECTOR_COND_MOVE (lx, lx + lx, cond2);
VECTOR_ELSE2 (cond2, cond)
VECTOR_IF2 ((hz | lz) == 0, cond3, cond2) // return sign(x)*0
VECTOR_RETURN (zeroes, cond3);
VECTOR_ENDIF
VECTOR_COND_MOVE (hx, hz + hz + (__builtin_convertvector(lz, v64usi) >> 31), cond2);
VECTOR_COND_MOVE (lx, lz + lz, cond2);
VECTOR_ENDIF
n += cond; // Active lanes should be -1
cond &= (n != 0);
}
hz = hx - hy;
lz = lx - ly;
VECTOR_COND_MOVE (hz, hz - 1, lx < ly);
VECTOR_IF (hz >= 0, cond)
VECTOR_COND_MOVE (hx, hz, cond);
VECTOR_COND_MOVE (lx, lz, cond);
VECTOR_ENDIF
/* convert back to floating value and restore the sign */
VECTOR_RETURN (zeroes, (hx | lx) == 0); // return sign(x)*0
cond = hx < 0x00100000;
while (!ALL_ZEROES_P (cond)) // normalize x
{
VECTOR_COND_MOVE (hx, hx + hx + (lx >> 31), cond);
VECTOR_COND_MOVE (lx, lx + lx, cond);
iy += cond; // Active lanes should be -1
cond &= (hx < 0x00100000);
}
VECTOR_IF (iy >= -1022, cond) // normalize output
VECTOR_COND_MOVE (hx, (hx - 0x00100000) | ((iy + 1023) << 20), cond);
INSERT_WORDS (x, hx | sx, lx, cond);
VECTOR_ELSE (cond) // subnormal output */
n = -1022 - iy;
VECTOR_IF2 (n <= 20, cond2, cond)
VECTOR_COND_MOVE (lx, (lx >> n) | (hx << (32 - n)), cond2);
VECTOR_COND_MOVE (hx, hx >> n, cond2);
VECTOR_ELSEIF2 (n <= 31, cond2, cond)
VECTOR_COND_MOVE (lx, __builtin_convertvector ((hx << (32 - n)) | (lx >> n), v64usi), cond2);
VECTOR_COND_MOVE (hx, sx, cond2);
VECTOR_ELSE2 (cond2, cond)
VECTOR_COND_MOVE (lx, __builtin_convertvector (hx >> (n - 32), v64usi), cond2);
VECTOR_COND_MOVE (hx, sx, cond2);
VECTOR_ENDIF
INSERT_WORDS (x, hx | sx, lx, cond);
x *= VECTOR_INIT (1.0); /* create necessary signal */
VECTOR_ENDIF
VECTOR_RETURN (x, NO_COND); /* exact output */
FUNCTION_RETURN;
}
DEF_VARIANTS2 (fmod, df, df)