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 | |
---|
22 | unsigned 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 | |
---|
130 | int ?\?( int x, unsigned int y ) { __CFA_EXP__(); } |
---|
131 | long int ?\?( long int x, unsigned long int y ) { __CFA_EXP__(); } |
---|
132 | long long int ?\?( long long int x, unsigned long long int y ) { __CFA_EXP__(); } |
---|
133 | unsigned int ?\?( unsigned int x, unsigned int y ) { __CFA_EXP__(); } |
---|
134 | unsigned long int ?\?( unsigned long int x, unsigned long int y ) { __CFA_EXP__(); } |
---|
135 | unsigned 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 | |
---|
140 | forall( 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__ |
---|