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
|
/*========================== 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;
} __satan_ha_c11 = {0xbc8917bfu};
static __constant union {
unsigned int w;
float f;
int i;
} __satan_ha_c10 = {0x3c0afad8u};
static __constant union {
unsigned int w;
float f;
int i;
} __satan_ha_c9 = {0x3cf6afd7u};
static __constant union {
unsigned int w;
float f;
int i;
} __satan_ha_c8 = {0xbd1e9009u};
static __constant union {
unsigned int w;
float f;
int i;
} __satan_ha_c7 = {0xbc0e2ce0u};
static __constant union {
unsigned int w;
float f;
int i;
} __satan_ha_c6 = {0x3d8ee88fu};
static __constant union {
unsigned int w;
float f;
int i;
} __satan_ha_c5 = {0xbd84d3d8u};
static __constant union {
unsigned int w;
float f;
int i;
} __satan_ha_c4 = {0xbd3a0aceu};
static __constant union {
unsigned int w;
float f;
int i;
} __satan_ha_c3 = {0x3e34898cu};
static __constant union {
unsigned int w;
float f;
int i;
} __satan_ha_c2 = {0xbe05f565u};
static __constant union {
unsigned int w;
float f;
int i;
} __satan_ha_c1l = {0x31b69212u};
static __constant union {
unsigned int w;
float f;
int i;
} __satan_ha_c1 = {0xbe8259aeu};
static __constant union {
unsigned int w;
float f;
int i;
} __satan_ha_c0 = {0xbd94e63fu};
__attribute__((always_inline)) inline int
__ocl_svml_internal_satan_ha(float *pxin, float *pres) {
int nRet = 0;
float xin = *pxin;
// float atanf_ha(float xin)
{
union {
unsigned int w;
float f;
int i;
} x, xa, hcorr, lcorr, ya, R0, R1, eps, Rcorr, sres;
int sgn_x, smask, sgn_r, diff;
float poly, R;
x.f = xin;
xa.w = x.w & 0x7fffffffu;
sgn_x = x.w ^ xa.w;
// y ~ 1/x
ya.f = 1.0f / (xa.f);
// smask = (|x|>1.0)? -1 : 0
diff = ya.w - xa.w;
smask = ((int)diff) >> 31;
// will compute pi/2 - atan(1/|x|) for |x|>1
hcorr.w = smask & 0xbfc90FDB;
lcorr.w = smask & 0x333BBD2E;
sgn_r = sgn_x ^ (smask & 0x80000000u);
// reciprocal relative error
eps.f = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )((-xa.f), ya.f, 1.0f);
// fixup for eps=NaN
eps.w &= 0xbfffffffu;
// reduced argument
R0.w = xa.w + (diff & smask);
// correction term for reduced argument
Rcorr.f = ya.f * eps.f;
// will not use Rcorr if |xa|<=|ya|
// fixup for Rcorr=NaN
Rcorr.w &= (smask & 0xbfffffffu);
R1.f = R0.f - 0.5f;
// add correction
R = R1.f + Rcorr.f;
poly = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(__satan_ha_c11.f, R,
__satan_ha_c10.f);
poly = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(poly, R, __satan_ha_c9.f);
poly = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(poly, R, __satan_ha_c8.f);
poly = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(poly, R, __satan_ha_c7.f);
poly = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(poly, R, __satan_ha_c6.f);
poly = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(poly, R, __satan_ha_c5.f);
poly = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(poly, R, __satan_ha_c4.f);
poly = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(poly, R, __satan_ha_c3.f);
poly = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(poly, R, __satan_ha_c2.f);
// _VSTATIC(c1l) can be used to reduce ulp error
// poly = SP_FMA(poly, R, _VSTATIC(c1l).f);
poly = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(poly, R, __satan_ha_c1.f);
poly = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(poly, R, __satan_ha_c0.f);
// used here when _VSTATIC(c1l) is used in poly evaluation
// poly = SP_FMA(_VSTATIC(c1).f, R, poly);
// Rcorr + Rcorr*poly
sres.f = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(poly, Rcorr.f, Rcorr.f);
// (R+ R*poly)+lcorr
sres.f = sres.f + lcorr.f;
// ((R+ R*poly)+lcorr)+R0*poly
sres.f = SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )(R0.f, poly, sres.f);
// ((R+ R*poly)+lcorr)+R0*poly + R0
sres.f = sres.f + R0.f;
// ((R+ R*poly)+lcorr)+Q0*poly + Q0 + hcorr
sres.f = sres.f + hcorr.f;
sres.w = sres.w ^ sgn_r;
*pres = sres.f;
}
return nRet;
}
float __ocl_svml_atanf_ha(float x) {
float r;
__ocl_svml_internal_satan_ha(&x, &r);
return r;
}
|
