sleef-sys 0.1.2

Rust FFI bindings to the SLEEF Vectorized Math Library 
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

//          Copyright Naoki Shibata 2010 - 2017.
// Distributed under the Boost Software License, Version 1.0.
//    (See accompanying file LICENSE.txt or copy at
//          http://www.boost.org/LICENSE_1_0.txt)

#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
#include <assert.h>
#include <math.h>
#include <time.h>
#include <unistd.h>
#include <x86intrin.h>

uint64_t Sleef_currentTimeMicros();
void fillDP(double *buf, double min, double max);
void fillSP(float *buf, double min, double max);

extern char x86BrandString[256], versionString[1024];
extern int veclen;
extern int enableLogExp;
extern double *abufdp, *bbufdp;
extern float *abufsp, *bbufsp;
extern FILE *fp;

#include "bench.h"

#ifdef __AVX512F__
typedef __m512d vdouble;
typedef __m512 vfloat;
#define ENABLED
#endif

#ifdef ENABLED
void benchSVML512_DPTrig() {
  fillDP(abufdp, 0, 6.28);

  callFuncSVML1_1(_mm512_sin_pd   , "sin, DP, 512", 0, 6.28, abufdp, vdouble);
  callFuncSVML1_1(_mm512_cos_pd   , "cos, DP, 512", 0, 6.28, abufdp, vdouble);
  callFuncSVML1_1(_mm512_tan_pd   , "tan, DP, 512", 0, 6.28, abufdp, vdouble);
  callFuncSVML2_1(_mm512_sincos_pd, "sincos, DP, 512", 0, 6.28, abufdp, vdouble);

  fillDP(abufdp, 0, 1e+6);
  
  callFuncSVML1_1(_mm512_sin_pd   , "sin, DP, 512", 0, 1e+6, abufdp, vdouble);
  callFuncSVML1_1(_mm512_cos_pd   , "cos, DP, 512", 0, 1e+6, abufdp, vdouble);
  callFuncSVML1_1(_mm512_tan_pd   , "tan, DP, 512", 0, 1e+6, abufdp, vdouble);
  callFuncSVML2_1(_mm512_sincos_pd, "sincos, DP, 512", 0, 1e+6, abufdp, vdouble);

  fillDP(abufdp, 0, 1e+100);
  
  callFuncSVML1_1(_mm512_sin_pd   , "sin, DP, 512", 0, 1e+100, abufdp, vdouble);
  callFuncSVML1_1(_mm512_cos_pd   , "cos, DP, 512", 0, 1e+100, abufdp, vdouble);
  callFuncSVML1_1(_mm512_tan_pd   , "tan, DP, 512", 0, 1e+100, abufdp, vdouble);
  callFuncSVML2_1(_mm512_sincos_pd, "sincos, DP, 512", 0, 1e+100, abufdp, vdouble);
}
  
void benchSVML512_DPNontrig() {
  fillDP(abufdp, 0, 1e+300);

  callFuncSVML1_1(_mm512_log_pd  , "log, DP, 512", 0, 1e+300, abufdp, vdouble);

  if (enableLogExp) {
    callFuncSVML1_1(_mm512_log10_pd, "log10, DP, 512", 0, 1e+300, abufdp, vdouble);
    callFuncSVML1_1(_mm512_log1p_pd, "log1p, DP, 512", 0, 1e+300, abufdp, vdouble);

    fillDP(abufdp, -700, 700);
  
    callFuncSVML1_1(_mm512_exp_pd  , "exp, DP, 512", -700, 700, abufdp, vdouble);
    callFuncSVML1_1(_mm512_exp2_pd , "exp2, DP, 512", -700, 700, abufdp, vdouble);
    callFuncSVML1_1(_mm512_exp10_pd, "exp10, DP, 512", -700, 700, abufdp, vdouble);

    fillDP(abufdp, -30, 30);
    fillDP(bbufdp, -30, 30);
  
    callFuncSVML1_2(_mm512_pow_pd, "pow, DP, 512", -30, 30, -30, 30, abufdp, bbufdp, vdouble);
  }
  
  fillDP(abufdp, -1.0, 1.0);
  
  callFuncSVML1_1(_mm512_asin_pd, "asin, DP, 512", -1.0, 1.0, abufdp, vdouble);
  callFuncSVML1_1(_mm512_acos_pd, "acos, DP, 512", -1.0, 1.0, abufdp, vdouble);

  fillDP(abufdp, -10, 10);
  fillDP(bbufdp, -10, 10);
  
  callFuncSVML1_1(_mm512_atan_pd, "atan, DP, 512", -10, 10, abufdp, vdouble);
  callFuncSVML1_2(_mm512_atan2_pd, "atan2, DP, 512", -10, 10, -10, 10, abufdp, bbufdp, vdouble);
}

void benchSVML512_SPTrig() {
  fillSP(abufsp, 0, 6.28);

  callFuncSVML1_1(_mm512_sin_ps   , "sin, SP, 512", 0, 6.28, abufsp, vfloat);
  callFuncSVML1_1(_mm512_cos_ps   , "cos, SP, 512", 0, 6.28, abufsp, vfloat);
  callFuncSVML1_1(_mm512_tan_ps   , "tan, SP, 512", 0, 6.28, abufsp, vfloat);
  callFuncSVML2_1(_mm512_sincos_ps, "sincos, SP, 512", 0, 6.28, abufsp, vfloat);

  fillSP(abufsp, 0, 1e+20);
  
  callFuncSVML1_1(_mm512_sin_ps   , "sin, SP, 512", 0, 1e+20, abufsp, vfloat);
  callFuncSVML1_1(_mm512_cos_ps   , "cos, SP, 512", 0, 1e+20, abufsp, vfloat);
  callFuncSVML1_1(_mm512_tan_ps   , "tan, SP, 512", 0, 1e+20, abufsp, vfloat);
  callFuncSVML2_1(_mm512_sincos_ps, "sincos, SP, 512", 0, 1e+20, abufsp, vfloat);
}

void benchSVML512_SPNontrig() {
  fillSP(abufsp, 0, 1e+38);

  callFuncSVML1_1(_mm512_log_ps  , "log, SP, 512", 0, 1e+38, abufsp, vfloat);

  if (enableLogExp) {
    callFuncSVML1_1(_mm512_log10_ps, "log10, SP, 512", 0, 1e+38, abufsp, vfloat);
    //callFuncSVML1_1(_mm512_log1p_ps, "log1p, SP, 512", 0, 1e+38, abufsp, vfloat);

    fillSP(abufsp, -100, 100);
  
    callFuncSVML1_1(_mm512_exp_ps  , "exp, SP, 512", -100, 100, abufsp, vfloat);
    callFuncSVML1_1(_mm512_exp2_ps , "exp2, SP, 512", -100, 100, abufsp, vfloat);
    callFuncSVML1_1(_mm512_exp10_ps, "exp10, SP, 512", -100, 100, abufsp, vfloat);

    fillSP(abufsp, -30, 30);
    fillSP(bbufsp, -30, 30);
  
    callFuncSVML1_2(_mm512_pow_ps, "pow, SP, 512", -30, 30, -30, 30, abufsp, bbufsp, vfloat);
  }
    
  fillSP(abufsp, -1.0, 1.0);
  
  callFuncSVML1_1(_mm512_asin_ps, "asin, SP, 512", -1.0, 1, abufsp, vfloat);
  callFuncSVML1_1(_mm512_acos_ps, "acos, SP, 512", -1.0, 1, abufsp, vfloat);

  fillSP(abufsp, -10, 10);
  fillSP(bbufsp, -10, 10);
  
  callFuncSVML1_1(_mm512_atan_ps, "atan, SP, 512", -10, 10, abufsp, vfloat);
  callFuncSVML1_2(_mm512_atan2_ps, "atan2, SP, 512", -10, 10, -10, 10, abufsp, bbufsp, vfloat);
}
#else // #ifdef ENABLED
void benchSVML512_DPTrig() {}
void benchSVML512_DPNontrig() {}
void benchSVML512_SPTrig() {}
void benchSVML512_SPNontrig() {}
#endif // #ifdef ENABLED