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
|
/**
* WHAM - high-throughput sequence aligner
* Copyright (C) 2011 WHAM Group, University of Wisconsin
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
/* $Id: model.cpp 165 2012-11-26 10:23:16Z yinan $ */
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include "model.h"
#include "error.h"
/*
* AlignerModel::computeNumIndex(int nError, int nPartition)
* Given the number of errors and the number of partitions,
* compute the number of required hash tables, according to the
* formula C(nPartition-1, nMismatch).
*/
int AlignerModel::computeNumIndex(int nError, int nPartition) {
int i;
unsigned long long x = 1;
for (i = nPartition - nError; i <= nPartition - 1; i++)
x *= i;
for (i = 1; i <= nError; i++)
x /= i;
return (int) x;
}
/*
* AlignerModel::computeNumLookup(int nError, int nPartition)
* Given the number of errors and the number of partitions,
* compute the number of lookups for each alignment, according
* to the formula C(nPartition, nMismatch).
*/
int AlignerModel::computeNumLookup(int nError, int nPartition) {
int i;
unsigned long long x = 1;
for (i = nPartition - nError + 1; i <= nPartition; i++)
x *= i;
for (i = 1; i <= nError; i++)
x /= i;
return (int) x;
}
/*
* AlignerModel::estimateNumPartition(int nEntry, int length, int nError)
* Estimate the optimal number of partitions based on a cost model, given
* parameters of the aligner.
*/
int AlignerModel::estimateNumPartition(unsigned int nEntry, int length,
int nError, bool memory) {
int p, bestp;
unsigned int nIndex, nLookup, nBucket;
double nSpace;
double costLookup, cost, min;
double p0, p1, p2;
min = 1000000000;
bestp = 0;
for (p = nError + 1; p <= nError + 5; p++) {
if (memory
&& estimateIndexSpace(nEntry, length, nError, p) >= getFreeMemory())
continue;
nIndex = computeNumIndex(nError, p);
nLookup = computeNumLookup(nError, p);
nSpace = pow(4.0, length / p * (p - nError));
nBucket = (double) nEntry < nSpace ? nEntry : (unsigned int) nSpace;
costLookup = 0;
/* probability of empty bucket */
p0 = pow(1 - double(1) / nBucket, nEntry);
costLookup += p0;
/* probability of non-overflow bucket */
p1 = nEntry * (double(1) / nBucket)
* pow(1 - double(1) / nBucket, nEntry - 1);
costLookup += 2 * p1;
/* probability of overflow bucket */
p2 = 1 - p0 - p1;
costLookup += p2 * ((nEntry - nEntry * p1) / (nBucket * p2));
cost = costLookup * nLookup;
if (cost < min) {
min = cost;
bestp = p;
}
}
return bestp;
}
bool AlignerModel::isFitMemory(unsigned int nEntry, int length, int nMismatch,
int nPartition) {
unsigned int szIndex;
unsigned int szMemory;
szIndex = estimateIndexSpace(nEntry, length, nMismatch, nPartition);
szMemory = getFreeMemory();
elog(DEBUG1, "Estimated index size: %d MB\n", szIndex);
elog(DEBUG1, "Free memory size: %d MB\n", szMemory);
/* left 200MB free space*/
return (szIndex + 200) < szMemory;
}
unsigned int AlignerModel::getNumHashtableFitMemory(unsigned int nEntry,
int length, int nMismatch, int nPartition) {
unsigned int szSeq;
unsigned int szIndex;
unsigned int szMemory;
unsigned int nHashtable;
szSeq = nEntry / 32 * 3 * sizeof(int) / 1024 / 1024;
nHashtable = computeNumIndex(nMismatch, nPartition);
szIndex = estimateHashtableSpace(nEntry, length, nMismatch, nPartition);
szMemory = getFreeMemory();
elog(DEBUG1, "Estimated hashtable size: %d MB\n", szIndex);
elog(DEBUG1, "Free memory size: %d MB\n", szMemory);
if ((szMemory - szSeq - 200) / szIndex < nHashtable)
return (szMemory - szSeq - 200) / szIndex;
else
return nHashtable;
}
unsigned int AlignerModel::getFreeMemory() {
unsigned int memTotal, memFree, memBuffer, memCache;
FILE * file;
// return 4096;
file = fopen("/proc/meminfo", "r");
if (file == NULL
) //on non-linux system
return 0;
fscanf(file, "MemTotal: %d kB\n", &memTotal);
fscanf(file, "MemFree: %d kB\n", &memFree);
fscanf(file, "Buffers: %d kB\n", &memBuffer);
fscanf(file, "Cached: %d kB\n", &memCache);
fclose(file);
return memTotal / 1024;
// return (memFree + memBuffer + memCache)/1024;
}
unsigned int AlignerModel::estimateIndexSpace(unsigned int nEntry, int length,
int nError, int nPartition) {
int nHashtable = 0;
unsigned int szSeq;
unsigned int szHashtable;
unsigned int szIndex;
szSeq = nEntry / 32 * 3 * sizeof(int) / 1024 / 1024;
nHashtable = computeNumIndex(nError, nPartition);
szHashtable = estimateHashtableSpace(nEntry, length, nError, nPartition);
szIndex = szSeq + szHashtable * nHashtable;
if (szIndex == 0)
szIndex = 1;
return szIndex;
}
unsigned int AlignerModel::estimateHashtableSpace(unsigned int nEntry,
int length, int nError, int nPartition) {
unsigned int szHashtable;
unsigned int nBucket;
double nSpace, p0, p1, p2;
nSpace = pow(4.0, length / nPartition * (nPartition - nError));
nBucket = (double) nEntry < nSpace ? nEntry : (unsigned int) nSpace;
/* probability of empty bucket */
p0 = pow(1 - double(1) / nBucket, nEntry);
/* probability of non-overflow bucket */
p1 = nEntry * (double(1) / nBucket)
* pow(1 - double(1) / nBucket, nEntry - 1);
/* probability of overflow bucket */
p2 = 1 - p0 - p1;
// nSpace = pow(8.0, length / nPartition * (nPartition - nError));
// nBucket = (double)nEntry < nSpace? nEntry : (unsigned int)nSpace;
szHashtable = (unsigned int) ((nEntry - nEntry * p1 + nBucket) / 1024 / 1024
* sizeof(int));
if (szHashtable == 0)
szHashtable = 1;
return szHashtable;
}
|
