1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
|
/*
* Copyright (c) 2014 Advanced Micro Devices, Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*/
#include <clc/clc.h>
#include <math/clc_remainder.h>
#include "../clcmacro.h"
#include "config.h"
#include "math.h"
_CLC_DEF _CLC_OVERLOAD float __clc_fmod(float x, float y)
{
int ux = as_int(x);
int ax = ux & EXSIGNBIT_SP32;
float xa = as_float(ax);
int sx = ux ^ ax;
int ex = ax >> EXPSHIFTBITS_SP32;
int uy = as_int(y);
int ay = uy & EXSIGNBIT_SP32;
float ya = as_float(ay);
int ey = ay >> EXPSHIFTBITS_SP32;
float xr = as_float(0x3f800000 | (ax & 0x007fffff));
float yr = as_float(0x3f800000 | (ay & 0x007fffff));
int c;
int k = ex - ey;
while (k > 0) {
c = xr >= yr;
xr -= c ? yr : 0.0f;
xr += xr;
--k;
}
c = xr >= yr;
xr -= c ? yr : 0.0f;
int lt = ex < ey;
xr = lt ? xa : xr;
yr = lt ? ya : yr;
float s = as_float(ey << EXPSHIFTBITS_SP32);
xr *= lt ? 1.0f : s;
c = ax == ay;
xr = c ? 0.0f : xr;
xr = as_float(sx ^ as_int(xr));
c = ax > PINFBITPATT_SP32 | ay > PINFBITPATT_SP32 | ax == PINFBITPATT_SP32 | ay == 0;
xr = c ? as_float(QNANBITPATT_SP32) : xr;
return xr;
}
_CLC_BINARY_VECTORIZE(_CLC_DEF _CLC_OVERLOAD, float, __clc_fmod, float, float);
#ifdef cl_khr_fp64
_CLC_DEF _CLC_OVERLOAD double __clc_fmod(double x, double y)
{
ulong ux = as_ulong(x);
ulong ax = ux & ~SIGNBIT_DP64;
ulong xsgn = ux ^ ax;
double dx = as_double(ax);
int xexp = convert_int(ax >> EXPSHIFTBITS_DP64);
int xexp1 = 11 - (int) clz(ax & MANTBITS_DP64);
xexp1 = xexp < 1 ? xexp1 : xexp;
ulong uy = as_ulong(y);
ulong ay = uy & ~SIGNBIT_DP64;
double dy = as_double(ay);
int yexp = convert_int(ay >> EXPSHIFTBITS_DP64);
int yexp1 = 11 - (int) clz(ay & MANTBITS_DP64);
yexp1 = yexp < 1 ? yexp1 : yexp;
// First assume |x| > |y|
// Set ntimes to the number of times we need to do a
// partial remainder. If the exponent of x is an exact multiple
// of 53 larger than the exponent of y, and the mantissa of x is
// less than the mantissa of y, ntimes will be one too large
// but it doesn't matter - it just means that we'll go round
// the loop below one extra time.
int ntimes = max(0, (xexp1 - yexp1) / 53);
double w = ldexp(dy, ntimes * 53);
w = ntimes == 0 ? dy : w;
double scale = ntimes == 0 ? 1.0 : 0x1.0p-53;
// Each time round the loop we compute a partial remainder.
// This is done by subtracting a large multiple of w
// from x each time, where w is a scaled up version of y.
// The subtraction must be performed exactly in quad
// precision, though the result at each stage can
// fit exactly in a double precision number.
int i;
double t, v, p, pp;
for (i = 0; i < ntimes; i++) {
// Compute integral multiplier
t = trunc(dx / w);
// Compute w * t in quad precision
p = w * t;
pp = fma(w, t, -p);
// Subtract w * t from dx
v = dx - p;
dx = v + (((dx - v) - p) - pp);
// If t was one too large, dx will be negative. Add back one w.
dx += dx < 0.0 ? w : 0.0;
// Scale w down by 2^(-53) for the next iteration
w *= scale;
}
// One more time
// Variable todd says whether the integer t is odd or not
t = floor(dx / w);
long lt = (long)t;
int todd = lt & 1;
p = w * t;
pp = fma(w, t, -p);
v = dx - p;
dx = v + (((dx - v) - p) - pp);
i = dx < 0.0;
todd ^= i;
dx += i ? w : 0.0;
// At this point, dx lies in the range [0,dy)
double ret = as_double(xsgn ^ as_ulong(dx));
dx = as_double(ax);
// Now handle |x| == |y|
int c = dx == dy;
t = as_double(xsgn);
ret = c ? t : ret;
// Next, handle |x| < |y|
c = dx < dy;
ret = c ? x : ret;
// We don't need anything special for |x| == 0
// |y| is 0
c = dy == 0.0;
ret = c ? as_double(QNANBITPATT_DP64) : ret;
// y is +-Inf, NaN
c = yexp > BIASEDEMAX_DP64;
t = y == y ? x : y;
ret = c ? t : ret;
// x is +=Inf, NaN
c = xexp > BIASEDEMAX_DP64;
ret = c ? as_double(QNANBITPATT_DP64) : ret;
return ret;
}
_CLC_BINARY_VECTORIZE(_CLC_DEF _CLC_OVERLOAD, double, __clc_fmod, double, double);
#endif
|
