source: libcfa/src/math.cfa @ e1358c0

Last change on this file since e1358c0 was f32448e, checked in by Andrew Beach <ajbeach@…>, 3 months ago

Fixed white-space. Woops.

  • Property mode set to 100644
File size: 6.0 KB
Line 
1//
2// Cforall Version 1.0.0 Copyright (C) 2015 University of Waterloo
3//
4// The contents of this file are covered under the licence agreement in the
5// file "LICENCE" distributed with Cforall.
6//
7// math.cpp --
8//
9// Author           : Andrew Beach
10// Created On       : Mon Nov 25 16:20:00 2024
11// Last Modified By : Andrew Beach
12// Created On       : Mon Nov 27 15:11:00 2024
13// Update Count     : 0
14//
15
16#include "math.hfa"
17
18#include <limits.h>
19
20#pragma GCC visibility push(default)
21
22unsigned long long log2_u32_32( unsigned long long val ) {
23        enum {
24                TABLE_BITS = 6,
25                TABLE_SIZE = (1 << TABLE_BITS) + 2,
26        };
27        // for(i; TABLE_SIZE) {
28        //  table[i] = (unsigned long long)(log2(1.0 + i / pow(2, TABLE_BITS)) * pow(2, 32)));
29        // }
30        static const unsigned long long table[] = {
31                0x0000000000, 0x0005b9e5a1, 0x000b5d69ba, 0x0010eb389f,
32                0x001663f6fa, 0x001bc84240, 0x002118b119, 0x002655d3c4,
33                0x002b803473, 0x00309857a0, 0x00359ebc5b, 0x003a93dc98,
34                0x003f782d72, 0x00444c1f6b, 0x0049101eac, 0x004dc4933a,
35                0x005269e12f, 0x00570068e7, 0x005b888736, 0x006002958c,
36                0x00646eea24, 0x0068cdd829, 0x006d1fafdc, 0x007164beb4,
37                0x00759d4f80, 0x0079c9aa87, 0x007dea15a3, 0x0081fed45c,
38                0x0086082806, 0x008a064fd5, 0x008df988f4, 0x0091e20ea1,
39                0x0095c01a39, 0x009993e355, 0x009d5d9fd5, 0x00a11d83f4,
40                0x00a4d3c25e, 0x00a8808c38, 0x00ac241134, 0x00afbe7fa0,
41                0x00b3500472, 0x00b6d8cb53, 0x00ba58feb2, 0x00bdd0c7c9,
42                0x00c1404ead, 0x00c4a7ba58, 0x00c80730b0, 0x00cb5ed695,
43                0x00ceaecfea, 0x00d1f73f9c, 0x00d53847ac, 0x00d8720935,
44                0x00dba4a47a, 0x00ded038e6, 0x00e1f4e517, 0x00e512c6e5,
45                0x00e829fb69, 0x00eb3a9f01, 0x00ee44cd59, 0x00f148a170,
46                0x00f446359b, 0x00f73da38d, 0x00fa2f045e, 0x00fd1a708b,
47                0x0100000000, 0x0102dfca16,
48        };
49        _Static_assert((sizeof(table) / sizeof(table[0])) == TABLE_SIZE, "TABLE_SIZE should be accurate");
50        // starting from val = (2 ** i)*(1 + f) where 0 <= f < 1
51        // log identities mean log2(val) = log2((2 ** i)*(1 + f)) = log2(2**i) + log2(1+f)
52        //
53        // getting i is easy to do using builtin_clz (count leading zero)
54        //
55        // we want to calculate log2(1+f) independently to have a many bits of precision as possible.
56        //     val = (2 ** i)*(1 + f) = 2 ** i   +   f * 2 ** i
57        // isolating f we get
58        //     val - 2 ** i = f * 2 ** i
59        //     (val - 2 ** i) / 2 ** i = f
60        //
61        // we want to interpolate from the table to get the values
62        // and compromise by doing quadratic interpolation (rather than higher degree interpolation)
63        //
64        // for the interpolation we want to shift everything the fist sample point
65        // so our parabola becomes x = 0
66        // this further simplifies the equations
67        //
68        // the consequence is that we need f in 2 forms:
69        //  - finding the index of x0
70        //  - finding the distance between f and x0
71        //
72        // since sample points are equidistant we can significantly simplify the equations
73
74        // get i
75        const unsigned long long bits = sizeof(val) * __CHAR_BIT__;
76        const unsigned long long lz = __builtin_clzl(val);
77        const unsigned long long i = bits - 1 - lz;
78
79        // get the fractinal part as a u32.32
80        const unsigned long long frac = (val << (lz + 1)) >> 32;
81
82        // get high order bits for the index into the table
83        const unsigned long long idx0 = frac >> (32 - TABLE_BITS);
84
85        // get the x offset, i.e., the difference between the first sample point and the actual fractional part
86        const long long udx = frac - (idx0 << (32 - TABLE_BITS));
87        /* paranoid */ verify((idx0 + 2) < TABLE_SIZE);
88
89        const long long y0 = table[idx0 + 0];
90        const long long y1 = table[idx0 + 1];
91        const long long y2 = table[idx0 + 2];
92
93        // from there we can quadraticly interpolate to get the data, using the lagrange polynomial
94        // normally it would look like:
95        //     double r0 = y0 * ((x - x1) / (x0 - x1)) * ((x - x2) / (x0 - x2));
96        //     double r1 = y1 * ((x - x0) / (x1 - x0)) * ((x - x2) / (x1 - x2));
97        //     double r2 = y2 * ((x - x0) / (x2 - x0)) * ((x - x1) / (x2 - x1));
98        // but since the spacing between sample points is fixed, we can simplify itand extract common expressions
99        const long long f1 = (y1 - y0);
100        const long long f2 = (y2 - y0);
101        const long long a = f2 - (f1 * 2l);
102        const long long b = (f1 * 2l) - a;
103
104        // Now we can compute it in the form (ax + b)x + c (which avoid repeating steps)
105        long long sum = ((a*udx) >> (32 - TABLE_BITS))  + b;
106        sum = (sum*udx) >> (32 - TABLE_BITS + 1);
107        sum = y0 + sum;
108
109        return (i << 32) + (sum);
110} // log2_u32_32
111
112// Implementation of power functions (from the prelude):
113
114#define __CFA_EXP__() \
115        if ( y == 0 ) return 1;                             /* convention */ \
116        __CFA_EXP_INT__(                                    /* special cases for integral types */ \
117                if ( x == 1 ) return 1;                         /* base case */ \
118                if ( x == 2 ) return x << (y - 1);              /* positive shifting */ \
119                if ( y >= sizeof(y) * CHAR_BIT ) return 0;      /* immediate overflow, negative exponent > 2^size-1 */ \
120        ) \
121        typeof(x) op = 1;                                   /* accumulate odd product */ \
122        typeof(x) w = x; /* FIX-ME: possible bug in the box pass changing value argument through parameter */ \
123        for ( ; y > 1; y >>= 1 ) {                          /* squaring exponentiation, O(log2 y) */ \
124                if ( (y & 1) == 1 ) op = op * w;                /* odd ? */ \
125                w = w * w; \
126        } \
127        return w * op
128#define __CFA_EXP_INT__(...) __VA_ARGS__
129
130int ?\?( int x, unsigned int y ) { __CFA_EXP__(); }
131long int ?\?( long int x, unsigned long int y ) { __CFA_EXP__(); }
132long long int ?\?( long long int x, unsigned long long int y ) { __CFA_EXP__(); }
133unsigned int ?\?( unsigned int x, unsigned int y ) { __CFA_EXP__(); }
134unsigned long int ?\?( unsigned long int x, unsigned long int y ) { __CFA_EXP__(); }
135unsigned long long int ?\?( unsigned long long int x, unsigned long long int y ) { __CFA_EXP__(); }
136
137#undef __CFA_EXP_INT__
138#define __CFA_EXP_INT__(...)
139
140forall( OT | { void ?{}( OT & this, one_t ); OT ?*?( OT, OT ); } ) {
141        OT ?\?( OT x, unsigned int y ) { __CFA_EXP__(); }
142        OT ?\?( OT x, unsigned long int y ) { __CFA_EXP__(); }
143        OT ?\?( OT x, unsigned long long int y ) { __CFA_EXP__(); }
144} // distribution
145
146#undef __CFA_EXP_INT__
147#undef __CFA_EXP__
Note: See TracBrowser for help on using the repository browser.