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
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
|
/*========================== begin_copyright_notice ============================
Copyright (C) 2017-2021 Intel Corporation
SPDX-License-Identifier: MIT
============================= end_copyright_notice ===========================*/
#include "../include/BiF_Definitions.cl"
#include "../../Headers/spirv.h"
#include "../IMF/FP32/sincos_s_la.cl"
#include "../IMF/FP32/sincos_s_noLUT.cl"
#include "../ExternalLibraries/libclc/trig.cl"
#if defined(cl_khr_fp64)
#include "../IMF/FP64/sincos_d_la.cl"
#include "../IMF/FP64/sincos_d_la_noLUT.cl"
#endif
static INLINE float __intel_sincos_f32_p0f32( float x, __private float* cosval, bool doFast )
{
float sin_x, cos_x;
if(__FastRelaxedMath && (!__APIRS) && doFast)
{
sin_x = SPIRV_OCL_BUILTIN(native_sin, _f32, )(x);
cos_x = SPIRV_OCL_BUILTIN(native_cos, _f32, )(x);
}
else if(__UseHighAccuracyMath)
{
sin_x = __ocl_svml_sincosf_noLUT(x, &cos_x);
}
else if(__UseMathWithLUT)
{
__ocl_svml_sincosf(x, &sin_x, &cos_x);
}
else
{
float abs_float = SPIRV_OCL_BUILTIN(fabs, _f32, )(x);
if( abs_float > 10000.0f )
{
sin_x = libclc_sin_f32(x);
cos_x = libclc_cos_f32(x);
}
else
{
sin_x = __ocl_svml_sincosf_noLUT(x, &cos_x);
}
}
*cosval = cos_x;
return sin_x;
}
INLINE float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(sincos, _f32_p0f32, )( float x, __private float* cosval )
{
return __intel_sincos_f32_p0f32(x, cosval, true);
}
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_1VAL_1PTRARG_LOOP( sincos, float, float, float, f32, f32 )
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(sincos, _f32_p1f32, )( float x,
__global float* cosval )
{
float sin_x, cos_x;
sin_x = SPIRV_OCL_BUILTIN(sincos, _f32_p0f32, )( x, &cos_x );
cosval[0] = cos_x;
return sin_x;
}
INLINE float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(sincos, _f32_p3f32, )( float x,
__local float* cosval )
{
float sin_x, cos_x;
sin_x = SPIRV_OCL_BUILTIN(sincos, _f32_p0f32, )( x, &cos_x );
cosval[0] = cos_x;
return sin_x;
}
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_1VALARG_1PTRARG( sincos, float, __global, float, f32, p1 )
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_1VALARG_1PTRARG( sincos, float, __local, float, f32, p3 )
#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
INLINE float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(sincos, _f32_p4f32, )( float x,
__generic float* cosval )
{
float sin_x, cos_x;
sin_x = SPIRV_OCL_BUILTIN(sincos, _f32_p0f32, )( x, &cos_x );
cosval[0] = cos_x;
return sin_x;
}
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_1VALARG_1PTRARG( sincos, float, __generic, float, f32, p4 )
#endif //#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
#if defined(cl_khr_fp16)
INLINE half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(sincos, _f16_p0f16, )( half x,
__private half* cosval )
{
float sin_x, cos_x;
sin_x = SPIRV_OCL_BUILTIN(sincos, _f32_p0f32, )( SPIRV_BUILTIN(FConvert, _f32_f16, _Rfloat)(x), &cos_x );
cosval[0] = SPIRV_BUILTIN(FConvert, _f16_f32, _Rhalf)(cos_x);
return SPIRV_BUILTIN(FConvert, _f16_f32, _Rhalf)(sin_x);
}
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_1VAL_1PTRARG_LOOP( sincos, half, half, half, f16, f16 )
INLINE half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(sincos, _f16_p1f16, )( half x,
__global half* cosval )
{
float sin_x, cos_x;
sin_x = SPIRV_OCL_BUILTIN(sincos, _f32_p0f32, )( SPIRV_BUILTIN(FConvert, _f32_f16, _Rfloat)(x), &cos_x );
cosval[0] = SPIRV_BUILTIN(FConvert, _f16_f32, _Rhalf)(cos_x);
return SPIRV_BUILTIN(FConvert, _f16_f32, _Rhalf)(sin_x);
}
INLINE half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(sincos, _f16_p3f16, )( half x,
__local half* cosval )
{
float sin_x, cos_x;
sin_x = SPIRV_OCL_BUILTIN(sincos, _f32_p0f32, )( SPIRV_BUILTIN(FConvert, _f32_f16, _Rfloat)(x), &cos_x );
cosval[0] = SPIRV_BUILTIN(FConvert, _f16_f32, _Rhalf)(cos_x);
return SPIRV_BUILTIN(FConvert, _f16_f32, _Rhalf)(sin_x);
}
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_1VALARG_1PTRARG( sincos, half, __global, half, f16, p1 )
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_1VALARG_1PTRARG( sincos, half, __local, half, f16, p3 )
#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
INLINE half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(sincos, _f16_p4f16, )( half x,
__generic half* cosval )
{
float sin_x, cos_x;
sin_x = SPIRV_OCL_BUILTIN(sincos, _f32_p0f32, )( SPIRV_BUILTIN(FConvert, _f32_f16, _Rfloat)(x), &cos_x );
cosval[0] = SPIRV_BUILTIN(FConvert, _f16_f32, _Rhalf)(cos_x);
return SPIRV_BUILTIN(FConvert, _f16_f32, _Rhalf)(sin_x);
}
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_1VALARG_1PTRARG( sincos, half, __generic, half, f16, p4 )
#endif //#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
#endif // defined(cl_khr_fp16)
#if defined(cl_khr_fp64)
INLINE double SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(sincos, _f64_p0f64, )( double x,
__private double* cosval )
{
double sin_x, cos_x;
if (__UseHighAccuracyMath) {
__ocl_svml_sincos_noLUT(x, &sin_x, &cos_x);
} else {
__ocl_svml_sincos(x, &sin_x, &cos_x);
}
*cosval = cos_x;
return sin_x;
}
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_1VAL_1PTRARG_LOOP( sincos, double, double, double, f64, f64 )
double SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(sincos, _f64_p3f64, )( double x,
__local double* cosval )
{
double sin_x, cos_x;
sin_x = SPIRV_OCL_BUILTIN(sincos, _f64_p0f64, )( x, &cos_x );
cosval[0] = cos_x;
return sin_x;
}
INLINE double SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(sincos, _f64_p1f64, )( double x,
__global double* cosval )
{
double sin_x, cos_x;
sin_x = SPIRV_OCL_BUILTIN(sincos, _f64_p0f64, )( x, &cos_x );
cosval[0] = cos_x;
return sin_x;
}
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_1VALARG_1PTRARG( sincos, double, __global, double, f64, p1 )
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_1VALARG_1PTRARG( sincos, double, __local, double, f64, p3 )
#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
INLINE double SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(sincos, _f64_p4f64, )( double x,
__generic double* cosval )
{
double sin_x, cos_x;
sin_x = SPIRV_OCL_BUILTIN(sincos, _f64_p0f64, )( x, &cos_x );
cosval[0] = cos_x;
return sin_x;
}
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_1VALARG_1PTRARG( sincos, double, __generic, double, f64, p4 )
#endif // (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
#endif // defined(cl_khr_fp64)
|
