-
Notifications
You must be signed in to change notification settings - Fork 2
/
iconEstimator.c
363 lines (296 loc) · 20.1 KB
/
iconEstimator.c
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
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
// Copyright 2018, Kevin Lang, Oath Research
#include <stdlib.h>
#include <stdio.h>
#include <assert.h>
#include <math.h>
#include "common.h"
#include "iconEstimator.h"
/******************************************************************************************/
// The ICON estimator for FM85 sketches is defined by the arXiv paper.
// The current file provides exact and approximate implementations of this estimator.
// The exact version works for any value of K, but is quite slow.
// The much faster approximate version works for K values that are powers of two
// ranging from 2^4 to 2^32.
// At a high-level, this approximation can be described as using an
// exponential approximation when C > K * (5.6 or 5.7), while smaller
// values of C are handled by a degree-19 polynomial approximation of
// a pre-conditioned version of the true ICON mapping from C to N_hat.
// This file also provides a validation procedure that compares its approximate
// and exact implementations of the FM85 ICON estimator.
/******************************************************************************************/
#define iconMinLogK 4
#define iconMaxLogK 26
// #define iconMaxLogK 32
#define iconPolynomialDegree 19
#define iconPolynomialNumCoefficients (1 + iconPolynomialDegree)
#define iconTableSize (iconPolynomialNumCoefficients * (1 + (iconMaxLogK - iconMinLogK)))
/******************************************************************************************/
double iconPolynomialCoefficents [iconTableSize] = {
// log K = 4
0.9895027971889700513, 0.3319496644645180128, 0.1242818722715769986, -0.03324149686026930256, -0.2985637298081619817,
1.366555923595830002, -4.705499366260569971, 11.61506432505530029, -21.11254986175579873, 28.89421695078809904,
-30.1383659011730991, 24.11946778830730054, -14.83391445199539938, 6.983088767267210173, -2.48964120264876998,
0.6593243603602499947, -0.125493534558034997, 0.01620971672896159843, -0.001271267679036929953, 4.567178653294529745e-05,
// log K = 5
0.9947713741300230339, 0.3326559581620939787, 0.1250050661634889981, -0.04130073804472530336, -0.2584095537451129854,
1.218050389433120051, -4.319106696095399656, 10.87175052045090062, -20.0184979022142997, 27.63210188163320069,
-28.97950009664030091, 23.26740804691930009, -14.33375703270860058, 6.751281271241110105, -2.406363094133439962,
0.6367414734718820357, -0.1210468076141379967, 0.01561196698118279963, -0.001222335432128580056, 4.383502970318410206e-05,
// log K = 6
0.9973904854982870161, 0.3330148852217920119, 0.125251536589509993, -0.04434075124043219962, -0.2436238890691720116,
1.163293254754570016, -4.177758779777369647, 10.60301981340099964, -19.6274507428828997, 27.18420839597660077,
-28.56827214174580121, 22.96268674086600114, -14.15234202220280046, 6.665700662642549901, -2.375043356720739851,
0.6280993991240929608, -0.119319019358031006, 0.01537674055733759954, -0.001202881695730769916, 4.309894633186929849e-05,
// log K = 7
0.9986963310058679655, 0.3331956705633329907, 0.125337696770523005, -0.04546817338088020299, -0.2386752211125199863,
1.145927328111949972, -4.135694445582720036, 10.52805060502839929, -19.52408322548339825, 27.06921653903929936,
-28.46207532143190022, 22.88083524357429965, -14.10057147392659971, 6.63958754983273991, -2.364865219283200037,
0.6251341806425250169, -0.1186991327450530043, 0.0152892726403408008, -0.001195439764873199896, 4.281098416794090072e-05,
// log K = 8
0.999348600452531044, 0.3332480372393080148, 0.126666900963325002, -0.06495714694254159371, -0.08376282050638980681,
0.3760158094643630267, -1.568204791601850001, 4.483117719555970382, -9.119180124379150598, 13.65799293358900002,
-15.3100211234349004, 12.97546344654869976, -8.351661538536939489, 4.075022612435580172, -1.49387015887069996,
0.4040976870253379927, -0.07813232681879349328, 0.01020545649538820085, -0.0008063279210812720381, 2.909334976414100078e-05,
// log K = 9
0.9996743787297059924, 0.3332925779481850093, 0.1267124599259649986, -0.06550452970936600228, -0.08191738117533520214,
0.3773034458363569987, -1.604679509609959975, 4.636761898691969641, -9.487348609558699408, 14.25164235443030059,
-15.99674955529870068, 13.56353219046370029, -8.730194904342459594, 4.259010067932120336, -1.56106689792022002,
0.4222540912786589828, -0.08165296504921559784, 0.01066878484925220041, -0.0008433887618256910015, 3.045339724886519912e-05,
// log K = 10
0.999837191783945034, 0.3333142252339619804, 0.1267759538087240012, -0.06631005632753710077, -0.07692759158286699428,
0.3568943956395980166, -1.546598721379510044, 4.51595019978557044, -9.298431968763770428, 14.02586858080080034,
-15.78858959520439953, 13.41484931677589998, -8.647958125130809748, 4.22398017468472009, -1.549708891200570093,
0.419507410264540026, -0.08117411611046250475, 0.01061202286184199928, -0.000839300527596772007, 3.03185874520205985e-05,
// log K = 11
0.9999186020796150265, 0.3333249054574359826, 0.126791713589799987, -0.06662487271699729652, -0.07335552427910230211,
0.3316370184815959909, -1.434143797561290068, 4.180260309967409604, -8.593906870708760692, 12.95088874800289958,
-14.56876092520539956, 12.37074367531410068, -7.969152075707960137, 3.888774396648960074, -1.424923326506990051,
0.385084561785229984, -0.07435541911616409816, 0.009695363567476529554, -0.0007644375960047160388, 2.75156194717188011e-05,
// log K = 12
0.9999592955649559967, 0.3333310560725140093, 0.1267379744020450116, -0.06524495415766619344, -0.08854031542298740343,
0.4244320628874230228, -1.794077789033230008, 5.133875262768450298, -10.40149374917120007, 15.47808115629240078,
-17.2272296137545986, 14.5002173676463002, -9.274819801602760094, 4.500782540026570189, -1.642359389030050076,
0.442596113445525019, -0.0853226219238850947, 0.01111969379054169975, -0.0008771614088006969611, 3.161668519459719752e-05,
// log K = 13
0.9999796468102559732, 0.3333336602394039727, 0.126728089053198989, -0.06503798598282370391, -0.09050261023823169548,
0.4350609244189960201, -1.831274835815670077, 5.223387516985289913, -10.55574395269979959, 15.67359470222429962,
-17.41263416341029924, 14.63297400889229927, -9.346752431221359458, 4.530124905188380069, -1.651245566462089975,
0.444542549250713015, -0.08561720963336499901, 0.01114805146185449992, -0.0008786251203363140043, 3.16416341644572998e-05,
// log K = 14
0.9999898187060970445, 0.3333362579300819806, 0.1266984078369459976, -0.06464561179765909715, -0.09343280886228019777,
0.4490702549264070087, -1.878087608052450008, 5.338004322057390283, -10.76690603590630069, 15.97069195083200022,
-17.73440379943459888, 14.90212518309260048, -9.520506013770420495, 4.616238931978830173, -1.68364817877918993,
0.4536194960681350086, -0.087448605434800597, 0.01139929991331390009, -0.0008995891451622229631, 3.244407259782900338e-05,
// log K = 15
0.9999949072549390028, 0.3333376334705290267, 0.126665364358402005, -0.06411790034705669439, -0.09776009134670660128,
0.4704691112248470253, -1.948021675295769972, 5.497760972696490001, -11.03165645315390009, 16.29703330781000048,
-18.03851029448010124, 15.11836776139680083, -9.638205179917429533, 4.665122328753120051, -1.698980686525759953,
0.4571799506245269873, -0.08804011353783609828, 0.01146553155965330043, -0.0009040455800659569869, 3.257931866957050274e-05,
// log K = 16
0.9999974544793589493, 0.3333381337614599871, 0.1266524862971120102, -0.06391676499117690535, -0.09929616211306059592,
0.4771390820378790254, -1.965762451227349938, 5.526802350376460282, -11.05703067024660058, 16.29535848023060041,
-18.00114005075790047, 15.06214012231560062, -9.58874727382628933, 4.63537541652793017, -1.686222848555620102,
0.4532602373715179933, -0.08719448925964939923, 0.01134365425717459921, -0.0008934965241274289835, 3.216436244471380105e-05,
// log K = 17
0.9999987278278800185, 0.3333383411464330148, 0.126642761751724009, -0.06371042959073920653, -0.1013564516034080043,
0.4891311195679299839, -2.010971712051409899, 5.644390807952309963, -11.27697253921500042, 16.59957157207080058,
-18.31808338317799922, 15.31363518393730061, -9.741451446816620674, 4.706207545519429658, -1.711102469010010063,
0.4597587341089349744, -0.08841670767182820134, 0.01149999225097850068, -0.0009056651366963050422, 3.259910736274500059e-05,
// log K = 18
0.9999993637727100371, 0.3333385511608860097, 0.1266341580529160016, -0.06353272828164230335, -0.103139962850642003,
0.4996216017206500104, -2.05099128585287982, 5.749874086531799655, -11.47727638570349917, 16.88141587810320132,
-18.61744656177490143, 15.55634230427719977, -9.892350736128680211, 4.778033520984200422, -1.737045483861280104,
0.4667410882683730167, -0.08977256212421590165, 0.01167940146667079994, -0.0009201381242396030127, 3.313600701586759867e-05,
// log K = 19
0.9999996805376010212, 0.3333372324328989778, 0.1267104737214659882, -0.06504749929326139601, -0.0882341962464350954,
0.4131871162041140244, -1.725190703567099915, 4.900817515593920426, -9.883452720776510603, 14.6657081190816001,
-16.29398295135089825, 13.69805011761319946, -8.753475239465899449, 4.244072374564439976, -1.547202527706629915,
0.4164770109614310267, -0.08017596922092029565, 0.01043146101701039954, -0.00082124200571200305, 2.953319493719429935e-05,
// log K = 20
0.9999998390037539986, 0.3333365859956040067, 0.1267460211029839967, -0.06569456024647769843, -0.0823070353477164951,
0.3810826463303410017, -1.611983580241109992, 4.624520077758210057, -9.397308335633589138, 14.03184981378050011,
-15.6703191315401007, 13.22992718704790072, -8.484216393184780713, 4.125607133488029987, -1.507690650697159906,
0.4066678517577320129, -0.07842110121777939868, 0.01021780862225150042, -0.0008054065857047439754, 2.899431830426989844e-05,
// log K = 21
0.9999999207001479817, 0.3333384953015239849, 0.1266331480396669928, -0.06345750166298599892, -0.1042341210992499961,
0.5077112908497130039, -2.087398133609810191, 5.858842546192500222, -11.70620319777190055, 17.23103975433669888,
-19.01462552846669851, 15.89674059836560005, -10.11395134034419918, 4.88760796465891989, -1.777886770904629987,
0.4780200178339499839, -0.09200895321782050218, 0.01198029553244219989, -0.0009447283875782100165, 3.405716775824710232e-05,
// log K = 22
0.9999999606908690497, 0.3333383929524300071, 0.1266456445096819927, -0.06373504294081690225, -0.1012834291081849969,
0.4893810690172959998, -2.01391428223606983, 5.656430437473649597, -11.3067201537791, 16.64980594135310099,
-18.3792355790383013, 15.36879753115040081, -9.778831246425049528, 4.725308061988969577, -1.718423596500280093,
0.4618308177809870019, -0.08883675060799739454, 0.01155766944804260087, -0.0009104695617243750358, 3.278237729674439666e-05,
// log K = 23
0.9999999794683379628, 0.3333386441751680085, 0.1266463995182049995, -0.06376031920455070556, -0.1010799540803130059,
0.488540137426137, -2.012048323537570127, 5.654949475342659682, -11.31023240892979942, 16.66334675284959843,
-18.40241452866079896, 15.39443572867130072, -9.798844412838670692, 4.736683907539640082, -1.723168363744929987,
0.463270349018644001, -0.08914619066708899531, 0.01160235936257320022, -0.0009143600818183229709, 3.293669304679140117e-05,
// log K = 24
0.9999999911469820146, 0.3333376076934529975, 0.1266944349940530012, -0.06470524278387919381, -0.09189342220283110152,
0.4359182372694809793, -1.815980282951169977, 5.149474056470340066, -10.37086570678100017, 15.36962686758569951,
-17.05756384717849983, 14.32755177515199918, -9.149944050025640152, 4.434601894497260055, -1.616478926806520056,
0.4351979157055039793, -0.08381768225272340223, 0.01091321820476520016, -0.0008600264403629039739, 3.09667800347144002e-05,
// log K = 25
0.9999999968592140354, 0.3333379164881000167, 0.1266782495827009913, -0.06434163088961859789, -0.09575258124988890451,
0.4597843575354370049, -1.911374431241559924, 5.411856661251520428, -10.88850084646090011, 16.12298941380269923,
-17.88172178487259956, 15.01301780636859995, -9.585542896142529301, 4.645811872761620442, -1.693952293156189892,
0.4563143308861309921, -0.08795976148455289523, 0.01146560428011200033, -0.0009048442931930629528, 3.26358391497329992e-05,
// log K = 26
0.9999999970700530483, 0.333338329556315982, 0.126644753076394001, -0.06372365346512399997, -0.1012760856945769949,
0.4886852278576360176, -2.009005418394389952, 5.638119224137019714, -11.26276715335160006, 16.57640024218650154,
-18.29035093605569884, 15.28892246224570073, -9.724916375991760731, 4.6978877652334603, -1.707974125916829955,
0.4588937864564729963, -0.08824617586088029375, 0.01147732114826570046, -0.00090384524860747295, 3.253252703695579795e-05,
#ifdef LARGER_K_VALUES
// log K = 27
1.000000000639100106, 0.3333378987508219815, 0.126670943746902992, -0.06418811974745139426, -0.0972951198506895043,
0.4687977077401049852, -1.945290489888900076, 5.499494964974400268, -11.05078190574979935, 16.3446428009706004,
-18.10936908931320133, 15.19089294103859977, -9.691829972777059155, 4.694320543263319934, -1.710719212277360013,
0.4606257962161550146, -0.08875858006645380438, 0.01156634964444109952, -0.0009125838337464230437, 3.290907977404550287e-05,
// log K = 28
0.9999999993590269476, 0.3333385660745579737, 0.1266394134278630013, -0.0636305053404186971, -0.1022354305220320031,
0.4945787360853979853, -2.032468917547570086, 5.702461924065530319, -11.38943406618639997, 16.76052144140630062,
-18.49169753114890113, 15.4564578116809006, -9.831507534599410292, 4.749667961030789698, -1.72701519749717991,
0.4640997252013580043, -0.08927103511252110213, 0.01161455495023329919, -0.000915030036039231982, 3.295110296010450275e-05,
// log K = 29
0.9999999998441060356, 0.3333383341194189886, 0.1266687338487519909, -0.06416245828383730643, -0.09764561286937140094,
0.4715274747139350242, -1.958172229464169911, 5.539587632966780362, -11.13784217611559946, 16.48149277721759987,
-18.26888916646990069, 15.33085193018819936, -9.78493991484172021, 4.741302923579859829, -1.728568959451310061,
0.4656457646521020011, -0.08977142058582450457, 0.01170492245846839995, -0.0009240931538567209464, 3.334703207098030245e-05,
// log K = 30
0.9999999992599339915, 0.3333384538468979752, 0.1266452025739940035, -0.06374775920488300052, -0.1009917742909720029,
0.4867931642504759737, -2.000981224888669807, 5.614968747087539569, -11.21527907219130071, 16.50500949673639894,
-18.21007853829650003, 15.22056128176249956, -9.680565515478869898, 4.675983737170599674, -1.69980511941418011,
0.4566332138743600111, -0.08779650251621799739, 0.01141656381272189956, -0.0008988545845624889468, 3.234448025291899689e-05,
// log K = 31
0.9999999973204000137, 0.333337762450663988, 0.1266965469104399944, -0.06475154253624139378, -0.09133098208494490333,
0.4320356889637699815, -1.799236887220760028, 5.100971076171499696, -10.27175516606700079, 15.22198757843720074,
-16.89368636262300072, 14.19016571851859965, -9.062390133299189188, 4.39220025249522994, -1.600994848692480099,
0.4310075283759189912, -0.08300339267288289746, 0.01080584419810979961, -0.0008514267355136160122, 3.065110087496039805e-05,
// log K = 32
0.9999999987706390536, 0.3333387038350890119, 0.1266354589419070031, -0.06355195838981600454, -0.102952771506954005,
0.4983589546197609854, -2.045281215270029929, 5.732181222451769642, -11.43849817800069957, 16.81961198331340057,
-18.54433120118400069, 15.49126422718470053, -9.84846998787154071, 4.755615082534379923, -1.728430514092559989,
0.4642927653670489985, -0.08927380119154580684, 0.01161055316485629964, -0.0009143724787632470305, 3.291492066818770055e-05,
#endif
};
/******************************************************************************************/
double evaluatePolynomial (double * coefficients, int start, int num, double x) {
int final = start + num - 1;
double total = coefficients[final];
int j;
for (j = final-1; j >= start; j--) {
total *= x;
total += coefficients[j];
}
return total;
}
/******************************************************************************************/
double iconExponentialApproximation (double k, double c) {
return (0.7940236163830469 * k * pow(2.0, c / k));
}
/******************************************************************************************/
double getIconEstimate (Short lgK, Long c) {
assert (lgK >= iconMinLogK);
assert (lgK <= iconMaxLogK);
if (c < 2L) return ((c == 0L) ? 0.0 : 1.0);
Long k = 1L << lgK;
double doubleK = (double) k;
double doubleC = (double) c;
// Differing thresholds ensure that the approximated estimator is monotonically increasing.
double thresholdFactor = ((lgK < 14) ? 5.7 : 5.6);
if (doubleC > (thresholdFactor * doubleK)) return (iconExponentialApproximation (doubleK, doubleC));
double factor = evaluatePolynomial (iconPolynomialCoefficents,
iconPolynomialNumCoefficients * (lgK - iconMinLogK),
iconPolynomialNumCoefficients,
// The somewhat arbitrary constant 2.0 is baked into the table iconPolynomialCoefficents[].
doubleC / (2.0 * doubleK));
double ratio = doubleC / doubleK;
// The somewhat arbitrary constant 66.774757 is baked into the table iconPolynomialCoefficents[].
double term = 1.0 + (ratio * ratio * ratio / 66.774757);
double result = doubleC * factor * term;
if (result >= doubleC) return result;
else return doubleC;
}
/******************************************************************************************/
/******************************************************************************************/
#ifdef SLOW_EXACT_VERSION
// Important note: do not change anything in the following function.
// It has been carefully designed and tested for numerical accuracy.
// In particular, the use of log1p and expm1 is critically important.
static inline double qnj (double kf, double nf, int col) {
double tmp1 = -1.0 / (kf * (pow (2.0, ((double) col))));
double tmp2 = log1p (tmp1);
return (-1.0 * (expm1 (nf * tmp2)));
}
double exactCofN (double kf, double nf) {
double total = 0.0;
int col;
for (col=128; col >= 1; col--) {
total += qnj (kf, nf, col);
};
return (kf * total);
}
#define iconInversionTolerance 1.0e-15
double exactIconEstimatorBinarySearch (double kf, double targetC, double nLoIn, double nHiIn) {
int depth = 0;
double nLo = nLoIn;
double nHi = nHiIn;
tailRecurse: // manual tail recursion optimization
if (depth > 100) {fprintf (stderr, "excessive recursion in binary search\n"); fflush (stderr); exit(EXIT_FAILURE);}
assert (nHi > nLo);
double nMid = nLo + 0.5 * (nHi - nLo);
assert (nMid > nLo && nMid < nHi);
if (((nHi - nLo) / nMid) < iconInversionTolerance) return (nMid);
double midC = exactCofN (kf, nMid);
if (midC == targetC) return (nMid);
if (midC < targetC) {nLo = nMid; depth++; goto tailRecurse;}
if (midC > targetC) {nHi = nMid; depth++; goto tailRecurse;}
{fprintf (stderr, "bad value in binary search\n"); fflush (stderr); exit(EXIT_FAILURE);}
}
double exactIconEstimatorBracketHi (double kf, double targetC, double nLo) {
int depth = 0;
double curN = 2.0 * nLo;
double curC = exactCofN(kf, curN);
while (curC <= targetC) {
if (depth > 100) {fprintf (stderr, "excessive looping in exactIconEstimatorBracketHi\n"); fflush (stderr); exit(EXIT_FAILURE);}
depth++;
curN *= 2.0;
curC = exactCofN(kf, curN);
}
assert (curC > targetC);
return (curN);
}
double exactIconEstimator (int lgK, long c) {
double targetC = (double) c;
if (c == 0L || c == 1L) return (targetC);
double kf = (double) (1L << lgK);
double nLo = targetC; assert (exactCofN(kf, nLo) < targetC); // bracket lo
double nHi = exactIconEstimatorBracketHi (kf, targetC, nLo); // bracket hi
return (exactIconEstimatorBinarySearch (kf, targetC, nLo, nHi));
}
/*
while (exactCofN(kf, nHi) <= targetC) {nHi *= 2.0;} // bracket
*/
#endif
/*******************************************************/
#ifdef TESTING_DRIVER
int main (int argc, char ** argv) {
if (argc != 2) {
fprintf (stderr, "Usage: %s lg_k > output_file\n", argv[0]); fflush (stderr);
exit(EXIT_FAILURE);
}
int lgK = atoi(argv[1]);
long k = (1L << lgK);
long c = 1;
while (c < k * 64) { // was k * 15
double exact = exactIconEstimator (lgK, c);
double approx = approximateIconEstimator (lgK, c);
double relDiff = (approx - exact) / exact;
printf ("%ld\t%.19g\t%.19g\t%.19g\n", c, relDiff, exact, approx);
long a = c+1;
long b = 1001 * c / 1000;
c = ((a > b) ? a : b);
}
}
#endif