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
|
/*
* 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 "config.h"
#include "../clcmacro.h"
#include "math.h"
_CLC_DEF _CLC_OVERLOAD float __clc_ldexp(float x, int n) {
if (!__clc_fp32_subnormals_supported()) {
// This treats subnormals as zeros
int i = as_int(x);
int e = (i >> 23) & 0xff;
int m = i & 0x007fffff;
int s = i & 0x80000000;
int v = add_sat(e, n);
v = clamp(v, 0, 0xff);
int mr = e == 0 | v == 0 | v == 0xff ? 0 : m;
int c = e == 0xff;
mr = c ? m : mr;
int er = c ? e : v;
er = e ? er : e;
return as_float( s | (er << 23) | mr );
}
/* supports denormal values */
const int multiplier = 24;
float val_f;
uint val_ui;
uint sign;
int exponent;
val_ui = as_uint(x);
sign = val_ui & 0x80000000;
val_ui = val_ui & 0x7fffffff;/* remove the sign bit */
int val_x = val_ui;
exponent = val_ui >> 23; /* get the exponent */
int dexp = exponent;
/* denormal support */
int fbh = 127 - (as_uint((float)(as_float(val_ui | 0x3f800000) - 1.0f)) >> 23);
int dexponent = 25 - fbh;
uint dval_ui = (( (val_ui << fbh) & 0x007fffff) | (dexponent << 23));
int ex = dexponent + n - multiplier;
dexponent = ex;
uint val = sign | (ex << 23) | (dval_ui & 0x007fffff);
int ex1 = dexponent + multiplier;
ex1 = -ex1 +25;
dval_ui = (((dval_ui & 0x007fffff )| 0x800000) >> ex1);
dval_ui = dexponent > 0 ? val :dval_ui;
dval_ui = dexponent > 254 ? 0x7f800000 :dval_ui; /*overflow*/
dval_ui = dexponent < -multiplier ? 0 : dval_ui; /*underflow*/
dval_ui = dval_ui | sign;
val_f = as_float(dval_ui);
exponent += n;
val = sign | (exponent << 23) | (val_ui & 0x007fffff);
ex1 = exponent + multiplier;
ex1 = -ex1 +25;
val_ui = (((val_ui & 0x007fffff )| 0x800000) >> ex1);
val_ui = exponent > 0 ? val :val_ui;
val_ui = exponent > 254 ? 0x7f800000 :val_ui; /*overflow*/
val_ui = exponent < -multiplier ? 0 : val_ui; /*underflow*/
val_ui = val_ui | sign;
val_ui = dexp == 0? dval_ui : val_ui;
val_f = as_float(val_ui);
val_f = isnan(x) | isinf(x) | val_x == 0 ? x : val_f;
return val_f;
}
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64 : enable
_CLC_DEF _CLC_OVERLOAD double __clc_ldexp(double x, int n) {
long l = as_ulong(x);
int e = (l >> 52) & 0x7ff;
long s = l & 0x8000000000000000;
ulong ux = as_ulong(x * 0x1.0p+53);
int de = ((int)(ux >> 52) & 0x7ff) - 53;
int c = e == 0;
e = c ? de: e;
ux = c ? ux : l;
int v = e + n;
v = clamp(v, -0x7ff, 0x7ff);
ux &= ~EXPBITS_DP64;
double mr = as_double(ux | ((ulong)(v+53) << 52));
mr = mr * 0x1.0p-53;
mr = v > 0 ? as_double(ux | ((ulong)v << 52)) : mr;
mr = v == 0x7ff ? as_double(s | PINFBITPATT_DP64) : mr;
mr = v < -53 ? as_double(s) : mr;
mr = ((n == 0) | isinf(x) | (x == 0) ) ? x : mr;
return mr;
}
#endif
#ifdef cl_khr_fp16
#pragma OPENCL EXTENSION cl_khr_fp16 : enable
_CLC_OVERLOAD _CLC_DEF half __clc_ldexp(half x, int n) {
return (half)__clc_ldexp((float)x, n);
}
_CLC_BINARY_VECTORIZE(_CLC_OVERLOAD _CLC_DEF, half, __clc_ldexp, half, int);
#endif
|
