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
|
/*========================== begin_copyright_notice ============================
Copyright (C) 2024 Intel Corporation
SPDX-License-Identifier: MIT
============================= end_copyright_notice ===========================*/
#include "../imf.h"
#pragma OPENCL FP_CONTRACT OFF
static __constant union {
unsigned int w;
float f;
int i;
} __dexp_ep_p_L2Ef = {0x3fB8AA3Bu};
static __constant union {
unsigned int w;
float f;
int i;
} __dexp_ep_p_Shifterf0 = {0x4b4003ffu};
static __constant union {
unsigned int w;
float f;
int i;
} __dexp_ep_p_fthres = {0x4431195c};
// -log(2)_high
static __constant union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} __dexp_ep_p_NL2H = {0xbfe62e42fefa39efuL};
// -log(2)_low
static __constant union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} __dexp_ep_p_NL2L = {0xbc7abc9e3b39803fuL};
static __constant union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} __dexp_ep_c6 = {0x3f56dd9818211af0uL};
static __constant union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} __dexp_ep_c5 = {0x3f8126fababd1cf2uL};
static __constant union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} __dexp_ep_c4 = {0x3fa55541c4c8cb89uL};
static __constant union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} __dexp_ep_c3 = {0x3fc55540432ea07buL};
static __constant union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} __dexp_ep_c2 = {0x3fe00000090aa64auL};
static __constant union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} __dexp_ep_c1 = {0x3ff000000a208385uL};
static __constant union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} __dexp_ep_c0 = {0xbdd63f26cce7780fuL};
static __constant union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} __dexp_ep_min_norm = {0x0010000000000000uL};
static __constant union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} __dexp_ep_Inf = {0x7ff0000000000000uL};
__attribute__((always_inline)) inline int
__ocl_svml_internal_dexp_ep(double *a, double *r) {
int nRet = 0;
double x = *a;
union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} xi, zero, res_special, scale;
union {
unsigned long w;
unsigned int w32[2];
int s32[2];
double f;
} T;
double N, R, R0, poly, res;
int expon32, mask32, mask_h;
unsigned int xa32, sgn_x, expon_corr;
union {
unsigned int w;
float f;
int i;
} idx;
float fN, xf;
xf = (float)x;
// x*log2(e) + Shifter
idx.f = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(xf, __dexp_ep_p_L2Ef.f,
__dexp_ep_p_Shifterf0.f);
// x*log2(e), rounded to integral
fN = idx.f - __dexp_ep_p_Shifterf0.f;
N = (double)fN;
// prepare exponent
T.w32[1] = idx.w << 20;
T.w32[0] = 0;
// reduced argument
R = SPIRV_OCL_BUILTIN(fma, _f64_f64_f64, )(__dexp_ep_p_NL2H.f, N, x);
// start polynomial computation
poly =
SPIRV_OCL_BUILTIN(fma, _f64_f64_f64, )(__dexp_ep_c6.f, R, __dexp_ep_c5.f);
poly = SPIRV_OCL_BUILTIN(fma, _f64_f64_f64, )(poly, R, __dexp_ep_c4.f);
poly = SPIRV_OCL_BUILTIN(fma, _f64_f64_f64, )(poly, R, __dexp_ep_c3.f);
poly = SPIRV_OCL_BUILTIN(fma, _f64_f64_f64, )(poly, R, __dexp_ep_c2.f);
poly = SPIRV_OCL_BUILTIN(fma, _f64_f64_f64, )(poly, R, __dexp_ep_c1.f);
poly = SPIRV_OCL_BUILTIN(fma, _f64_f64_f64, )(poly, R, __dexp_ep_c0.f);
if (SPIRV_OCL_BUILTIN(fabs, _f32, )(xf) > __dexp_ep_p_fthres.f)
goto EXP_SPECIAL_PATH;
// result
res = SPIRV_OCL_BUILTIN(fma, _f64_f64_f64, )(T.f, poly, T.f);
*r = res;
return nRet;
EXP_SPECIAL_PATH:
xi.f = x;
xa32 = xi.w32[1] & 0x7fffffffu;
// sign of x
sgn_x = xa32 ^ xi.w32[1];
if (xa32 < 0x40879127u) {
expon_corr = sgn_x ? 0x08000000u : 0xF8000000u;
scale.w = sgn_x ? 0x37f0000000000000uL : 0x47f0000000000000uL;
// apply correction (+/-128) to exponent embedded in T
T.w32[1] += expon_corr;
// result
res = SPIRV_OCL_BUILTIN(fma, _f64_f64_f64, )(T.f, poly, T.f);
// final scaling
res *= scale.f;
} else {
// underflow or overflow?
res_special.w = sgn_x ? 0x0000000000000000uL : 0x7ff0000000000000uL;
// check for NaNs
xi.w32[1] = xa32;
res_special.f = (xi.w <= 0x7ff0000000000000uL) ? res_special.f : x;
// quietize NaNs
zero.w = 0;
res = res_special.f + zero.f;
}
nRet = (res < __dexp_ep_min_norm.f) ? 4 : nRet;
nRet = (res == __dexp_ep_Inf.f) ? 3 : nRet;
*r = res;
return nRet;
}
double __ocl_svml_exp_ep(double x) {
double r;
__ocl_svml_internal_dexp_ep(&x, &r);
return r;
}
|
