line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
1
|
|
|
|
|
|
|
#define PERL_NO_GET_CONTEXT |
2
|
|
|
|
|
|
|
#define NO_XSLOCKS |
3
|
|
|
|
|
|
|
#include "EXTERN.h" |
4
|
|
|
|
|
|
|
#include "perl.h" |
5
|
|
|
|
|
|
|
#include "XSUB.h" |
6
|
|
|
|
|
|
|
|
7
|
|
|
|
|
|
|
#include "ppport.h" |
8
|
|
|
|
|
|
|
|
9
|
|
|
|
|
|
|
#define NUM_LATTICES 4 |
10
|
|
|
|
|
|
|
|
11
|
|
|
|
|
|
|
/* |
12
|
|
|
|
|
|
|
* This program must deal with integers that are too big to be |
13
|
|
|
|
|
|
|
* represented by 32 bits. |
14
|
|
|
|
|
|
|
* |
15
|
|
|
|
|
|
|
* They are represented by AM_BIG_INT, which is typedef'd to |
16
|
|
|
|
|
|
|
* |
17
|
|
|
|
|
|
|
* unsigned long a[8] |
18
|
|
|
|
|
|
|
* |
19
|
|
|
|
|
|
|
* where each a[i] < 2*16. Such an array represents the integer |
20
|
|
|
|
|
|
|
* |
21
|
|
|
|
|
|
|
* a[0] + a[1] * 2^16 + ... + a[7] * 2^(7*16). |
22
|
|
|
|
|
|
|
* |
23
|
|
|
|
|
|
|
* We only use 16 bits of the unsigned long instead of 32, so that |
24
|
|
|
|
|
|
|
* when we add or multiply two large integers, we have room for overflow. |
25
|
|
|
|
|
|
|
* After any addition or multiplication, the result is carried so that |
26
|
|
|
|
|
|
|
* each element of the array is again < 2*16. |
27
|
|
|
|
|
|
|
* |
28
|
|
|
|
|
|
|
* Someday I may rewrite this in assembler. |
29
|
|
|
|
|
|
|
* |
30
|
|
|
|
|
|
|
*/ |
31
|
|
|
|
|
|
|
typedef unsigned short AM_SHORT; |
32
|
|
|
|
|
|
|
typedef unsigned long AM_LONG; |
33
|
|
|
|
|
|
|
typedef AM_LONG AM_BIG_INT[8]; |
34
|
|
|
|
|
|
|
|
35
|
|
|
|
|
|
|
#define high_bits(x) x >> 16 |
36
|
|
|
|
|
|
|
#define low_bits(x) x & 0xffff |
37
|
|
|
|
|
|
|
|
38
|
|
|
|
|
|
|
#define carry(var, ind) \ |
39
|
|
|
|
|
|
|
var[ind + 1] += high_bits(var[ind]); \ |
40
|
|
|
|
|
|
|
var[ind] = low_bits(var[ind]) |
41
|
|
|
|
|
|
|
|
42
|
|
|
|
|
|
|
/* carry macro for AM_BIG_INT pointers */ |
43
|
|
|
|
|
|
|
#define carry_pointer(p) \ |
44
|
|
|
|
|
|
|
*(p + 1) += high_bits(*(p)); \ |
45
|
|
|
|
|
|
|
*(p) = low_bits(*(p)) |
46
|
|
|
|
|
|
|
|
47
|
|
|
|
|
|
|
#define carry_replace(var, ind) \ |
48
|
|
|
|
|
|
|
var[ind + 1] = high_bits(var[ind]); \ |
49
|
|
|
|
|
|
|
var[ind] = low_bits(var[ind]) |
50
|
|
|
|
|
|
|
|
51
|
|
|
|
|
|
|
/* |
52
|
|
|
|
|
|
|
* structure for the supracontexts |
53
|
|
|
|
|
|
|
* |
54
|
|
|
|
|
|
|
*/ |
55
|
|
|
|
|
|
|
|
56
|
|
|
|
|
|
|
typedef struct AM_supra { |
57
|
|
|
|
|
|
|
/* list of subcontexts in this supracontext |
58
|
|
|
|
|
|
|
* |
59
|
|
|
|
|
|
|
* data[0] is the number of subcontexts in |
60
|
|
|
|
|
|
|
* the array; |
61
|
|
|
|
|
|
|
* |
62
|
|
|
|
|
|
|
* data[1] is always 0 (useful for finding |
63
|
|
|
|
|
|
|
* intersections; see below) |
64
|
|
|
|
|
|
|
* |
65
|
|
|
|
|
|
|
* data[i] is not an actually subcontext |
66
|
|
|
|
|
|
|
* label; instead, all the subcontext labels |
67
|
|
|
|
|
|
|
* are kept in an array called subcontext |
68
|
|
|
|
|
|
|
* (bad choice of name?) created in |
69
|
|
|
|
|
|
|
* function _fillandcount(). Thus, the |
70
|
|
|
|
|
|
|
* actual subcontexts in the supracontext |
71
|
|
|
|
|
|
|
* are subcontext[data[2]], ... |
72
|
|
|
|
|
|
|
* |
73
|
|
|
|
|
|
|
* data[i] < data[i+1] if i > 1 and |
74
|
|
|
|
|
|
|
* i < data[0]. |
75
|
|
|
|
|
|
|
* |
76
|
|
|
|
|
|
|
* Using an array of increasing positive |
77
|
|
|
|
|
|
|
* integers makes it easy to take |
78
|
|
|
|
|
|
|
* intersections (see lattice.pod). |
79
|
|
|
|
|
|
|
*/ |
80
|
|
|
|
|
|
|
AM_SHORT *data; |
81
|
|
|
|
|
|
|
|
82
|
|
|
|
|
|
|
/* number of supracontexts that contain |
83
|
|
|
|
|
|
|
* precisely these subcontexts; |
84
|
|
|
|
|
|
|
* |
85
|
|
|
|
|
|
|
* According to the AM algorithm, we're |
86
|
|
|
|
|
|
|
* supposed to look at all the homogeneous |
87
|
|
|
|
|
|
|
* supracontexts to compute the analogical |
88
|
|
|
|
|
|
|
* set. Instead of traversing the |
89
|
|
|
|
|
|
|
* supracontextual lattice to find them, we |
90
|
|
|
|
|
|
|
* can instead traverse the list of AM_SUPRA |
91
|
|
|
|
|
|
|
* with count > 0 and use the value of count |
92
|
|
|
|
|
|
|
* to do our computing. |
93
|
|
|
|
|
|
|
* |
94
|
|
|
|
|
|
|
* Since we're actually traversing four |
95
|
|
|
|
|
|
|
* small lattices and taking intersections, |
96
|
|
|
|
|
|
|
* we'll be multiplying the four values of |
97
|
|
|
|
|
|
|
* count to get what we want. |
98
|
|
|
|
|
|
|
* |
99
|
|
|
|
|
|
|
*/ |
100
|
|
|
|
|
|
|
AM_SHORT count; |
101
|
|
|
|
|
|
|
|
102
|
|
|
|
|
|
|
/* |
103
|
|
|
|
|
|
|
* used to implement two linked lists |
104
|
|
|
|
|
|
|
* |
105
|
|
|
|
|
|
|
* One linked list contains all the nonempty |
106
|
|
|
|
|
|
|
* supracontexts (i.e., data[0] is not 0). |
107
|
|
|
|
|
|
|
* This linked list is in fact circular. |
108
|
|
|
|
|
|
|
* |
109
|
|
|
|
|
|
|
* One linked list contains all the unused |
110
|
|
|
|
|
|
|
* memory that can be used for new |
111
|
|
|
|
|
|
|
* supracontexts. |
112
|
|
|
|
|
|
|
*/ |
113
|
|
|
|
|
|
|
AM_SHORT next; |
114
|
|
|
|
|
|
|
|
115
|
|
|
|
|
|
|
/* |
116
|
|
|
|
|
|
|
* used during the filling of the |
117
|
|
|
|
|
|
|
* supracontextual lattice (see below) |
118
|
|
|
|
|
|
|
*/ |
119
|
|
|
|
|
|
|
unsigned char touched; |
120
|
|
|
|
|
|
|
} AM_SUPRA; |
121
|
|
|
|
|
|
|
|
122
|
|
|
|
|
|
|
/* |
123
|
|
|
|
|
|
|
* There is quite a bit of data that must pass between AM.pm and |
124
|
|
|
|
|
|
|
* AM.xs. Instead of repeatedly passing it back and forth on |
125
|
|
|
|
|
|
|
* the argument stack, AM.pm sends references to the variables |
126
|
|
|
|
|
|
|
* holding this shared data, by calling _xs_initialize() (defined later |
127
|
|
|
|
|
|
|
* on). These pointers are then stored in the following structure, |
128
|
|
|
|
|
|
|
* which is put into the magic part of $self (since $self is an HV, |
129
|
|
|
|
|
|
|
* it is perforce an SvPVMG as well). |
130
|
|
|
|
|
|
|
* |
131
|
|
|
|
|
|
|
* Note that for arrays, we store a pointer to the array data itself, |
132
|
|
|
|
|
|
|
* not the AV*. That means that in AM.pm, we have to be careful |
133
|
|
|
|
|
|
|
* how we make assignments to array variables; a reassignment such as |
134
|
|
|
|
|
|
|
* |
135
|
|
|
|
|
|
|
* @sum = (pack "L!8", 0, 0, 0, 0, 0, 0, 0, 0) x @sum; |
136
|
|
|
|
|
|
|
* |
137
|
|
|
|
|
|
|
* breaks everything because the pointer stored here then won't point |
138
|
|
|
|
|
|
|
* to the actual data anymore. That's why the appropriate line in |
139
|
|
|
|
|
|
|
* AM.pm is |
140
|
|
|
|
|
|
|
* |
141
|
|
|
|
|
|
|
* foreach (@sum) { |
142
|
|
|
|
|
|
|
* $_ = pack "L!8", 0, 0, 0, 0, 0, 0, 0, 0; |
143
|
|
|
|
|
|
|
* } |
144
|
|
|
|
|
|
|
* |
145
|
|
|
|
|
|
|
* Most of the identifiers in the struct have the same names as the |
146
|
|
|
|
|
|
|
* variables created in AM.pm and are documented there. Those |
147
|
|
|
|
|
|
|
* that don't are documented below. |
148
|
|
|
|
|
|
|
* |
149
|
|
|
|
|
|
|
* This trick of storing pointers like this is borrowed from the |
150
|
|
|
|
|
|
|
* source code of Perl/Tk. Thanks, Nick! |
151
|
|
|
|
|
|
|
* |
152
|
|
|
|
|
|
|
*/ |
153
|
|
|
|
|
|
|
|
154
|
|
|
|
|
|
|
typedef struct AM_guts { |
155
|
|
|
|
|
|
|
|
156
|
|
|
|
|
|
|
/* |
157
|
|
|
|
|
|
|
* Let i be an integer from 0 to 3; this represents which of the |
158
|
|
|
|
|
|
|
* four sublattices we are considering. |
159
|
|
|
|
|
|
|
* |
160
|
|
|
|
|
|
|
* Let lattice = lptr[i] and supralist = sptr[i]; then lattice and |
161
|
|
|
|
|
|
|
* supralist taken together tell us which subcontexts are in a |
162
|
|
|
|
|
|
|
* particular supracontext. If s is the label of a supracontext, |
163
|
|
|
|
|
|
|
* then it contains the subcontexts listed in |
164
|
|
|
|
|
|
|
* supralist[lattice[s]].data[]. |
165
|
|
|
|
|
|
|
* |
166
|
|
|
|
|
|
|
*/ |
167
|
|
|
|
|
|
|
|
168
|
|
|
|
|
|
|
AM_SHORT *lptr[NUM_LATTICES]; |
169
|
|
|
|
|
|
|
AM_SUPRA *sptr[NUM_LATTICES]; |
170
|
|
|
|
|
|
|
|
171
|
|
|
|
|
|
|
/* array ref containing number of active features in |
172
|
|
|
|
|
|
|
* each lattice (currently we us four lattices) |
173
|
|
|
|
|
|
|
*/ |
174
|
|
|
|
|
|
|
SV **lattice_sizes; |
175
|
|
|
|
|
|
|
/* array ref containing class labels for whole data set; |
176
|
|
|
|
|
|
|
* array index is data item index in data set. |
177
|
|
|
|
|
|
|
*/ |
178
|
|
|
|
|
|
|
SV **classes; |
179
|
|
|
|
|
|
|
/* ??? */ |
180
|
|
|
|
|
|
|
SV **itemcontextchain; |
181
|
|
|
|
|
|
|
/* ??? */ |
182
|
|
|
|
|
|
|
HV *itemcontextchainhead; |
183
|
|
|
|
|
|
|
/* Maps subcontext binary labels to class indices */ |
184
|
|
|
|
|
|
|
HV *context_to_class; |
185
|
|
|
|
|
|
|
/* Maps binary context labels to the number of training items |
186
|
|
|
|
|
|
|
* contained in that subcontext |
187
|
|
|
|
|
|
|
*/ |
188
|
|
|
|
|
|
|
HV *contextsize; |
189
|
|
|
|
|
|
|
/* Maps binary context labels to the number of pointers to each, |
190
|
|
|
|
|
|
|
* or to the number of pointers to class label if heterogenous. |
191
|
|
|
|
|
|
|
* The key 'grandtotal' maps to the total number of pointers. |
192
|
|
|
|
|
|
|
*/ |
193
|
|
|
|
|
|
|
HV *pointers; |
194
|
|
|
|
|
|
|
/* Maps binary context labels to the size of the gang effect of |
195
|
|
|
|
|
|
|
* that context. A gang effect is the number of pointers in |
196
|
|
|
|
|
|
|
* the given context multiplied by the number training items |
197
|
|
|
|
|
|
|
* contained in the context. |
198
|
|
|
|
|
|
|
*/ |
199
|
|
|
|
|
|
|
HV *gang; |
200
|
|
|
|
|
|
|
/* number of pointers to each class label; |
201
|
|
|
|
|
|
|
* keys are class indices and values are numbers |
202
|
|
|
|
|
|
|
* of pointers (AM_BIG_INT). |
203
|
|
|
|
|
|
|
*/ |
204
|
|
|
|
|
|
|
SV **sum; |
205
|
|
|
|
|
|
|
/* |
206
|
|
|
|
|
|
|
* contains the total number of possible class labels; |
207
|
|
|
|
|
|
|
* used for computing gang effects. |
208
|
|
|
|
|
|
|
*/ |
209
|
|
|
|
|
|
|
IV num_classes; |
210
|
|
|
|
|
|
|
} AM_GUTS; |
211
|
|
|
|
|
|
|
|
212
|
|
|
|
|
|
|
/* |
213
|
|
|
|
|
|
|
* A function and a vtable necessary for the use of Perl magic |
214
|
|
|
|
|
|
|
*/ |
215
|
|
|
|
|
|
|
|
216
|
192
|
|
|
|
|
|
static int AMguts_mgFree(pTHX_ SV *sv, MAGIC *mg) { |
217
|
|
|
|
|
|
|
int i; |
218
|
192
|
|
|
|
|
|
AM_GUTS *guts = (AM_GUTS *) SvPVX(mg->mg_obj); |
219
|
960
|
100
|
|
|
|
|
for (i = 0; i < NUM_LATTICES; ++i) { |
220
|
768
|
|
|
|
|
|
Safefree(guts->lptr[i]); |
221
|
768
|
|
|
|
|
|
Safefree(guts->sptr[i][0].data); |
222
|
768
|
|
|
|
|
|
Safefree(guts->sptr[i]); |
223
|
|
|
|
|
|
|
} |
224
|
192
|
|
|
|
|
|
return 0; |
225
|
|
|
|
|
|
|
} |
226
|
|
|
|
|
|
|
|
227
|
|
|
|
|
|
|
MGVTBL AMguts_vtab = { |
228
|
|
|
|
|
|
|
NULL, |
229
|
|
|
|
|
|
|
NULL, |
230
|
|
|
|
|
|
|
NULL, |
231
|
|
|
|
|
|
|
NULL, |
232
|
|
|
|
|
|
|
AMguts_mgFree |
233
|
|
|
|
|
|
|
}; |
234
|
|
|
|
|
|
|
|
235
|
|
|
|
|
|
|
/* |
236
|
|
|
|
|
|
|
* arrays used in the change-of-base portion of normalize(SV *s) |
237
|
|
|
|
|
|
|
* they are initialized in BOOT |
238
|
|
|
|
|
|
|
* |
239
|
|
|
|
|
|
|
*/ |
240
|
|
|
|
|
|
|
|
241
|
|
|
|
|
|
|
AM_LONG tens[16]; /* 10, 10*2, 10*4, ... */ |
242
|
|
|
|
|
|
|
AM_LONG ones[16]; /* 1, 1*2, 1*4, ... */ |
243
|
|
|
|
|
|
|
|
244
|
|
|
|
|
|
|
/* |
245
|
|
|
|
|
|
|
* function: normalize(SV *s) |
246
|
|
|
|
|
|
|
* |
247
|
|
|
|
|
|
|
* s is an SvPV whose PV* is a unsigned long array representing a very |
248
|
|
|
|
|
|
|
* large integer |
249
|
|
|
|
|
|
|
* |
250
|
|
|
|
|
|
|
* this function modifies s so that its NV is the floating point |
251
|
|
|
|
|
|
|
* representation of the very large integer value, while its PV* is |
252
|
|
|
|
|
|
|
* the decimal representation of the very large integer value in ASCII |
253
|
|
|
|
|
|
|
* (cool, a double-valued scalar) |
254
|
|
|
|
|
|
|
* |
255
|
|
|
|
|
|
|
* computing the NV is straightforward |
256
|
|
|
|
|
|
|
* |
257
|
|
|
|
|
|
|
* computing the PV is done using the old change-of-base algorithm: |
258
|
|
|
|
|
|
|
* repeatedly divide by 10, and use the remainders to construct the |
259
|
|
|
|
|
|
|
* ASCII digits from least to most significant |
260
|
|
|
|
|
|
|
* |
261
|
|
|
|
|
|
|
*/ |
262
|
|
|
|
|
|
|
|
263
|
6243
|
|
|
|
|
|
normalize(pTHX_ SV *s) { |
264
|
|
|
|
|
|
|
AM_LONG dspace[10]; |
265
|
|
|
|
|
|
|
AM_LONG qspace[10]; |
266
|
|
|
|
|
|
|
char outspace[55]; |
267
|
|
|
|
|
|
|
AM_LONG *dividend, *quotient, *dptr, *qptr; |
268
|
|
|
|
|
|
|
char *outptr; |
269
|
6243
|
|
|
|
|
|
unsigned int outlength = 0; |
270
|
6243
|
|
|
|
|
|
AM_LONG *p = (AM_LONG *) SvPVX(s); |
271
|
6243
|
|
|
|
|
|
STRLEN length = SvCUR(s) / sizeof(AM_LONG); |
272
|
|
|
|
|
|
|
/* TODO: is this required to be a certain number of bits?*/ |
273
|
6243
|
|
|
|
|
|
long double nn = 0; |
274
|
|
|
|
|
|
|
int j; |
275
|
|
|
|
|
|
|
|
276
|
|
|
|
|
|
|
/* you can't put the for block in {}, or it doesn't work |
277
|
|
|
|
|
|
|
* ask me for details some time |
278
|
|
|
|
|
|
|
* TODO: is this still necessary? (Nate) |
279
|
|
|
|
|
|
|
*/ |
280
|
56187
|
100
|
|
|
|
|
for (j = 8; j; --j){ |
281
|
|
|
|
|
|
|
/* 2^16 * nn + p[j-1] */ |
282
|
49944
|
|
|
|
|
|
nn = 65536.0 * nn + (double) *(p + j - 1); |
283
|
|
|
|
|
|
|
} |
284
|
|
|
|
|
|
|
|
285
|
6243
|
|
|
|
|
|
dividend = &dspace[0]; |
286
|
6243
|
|
|
|
|
|
quotient = &qspace[0]; |
287
|
6243
|
50
|
|
|
|
|
Copy(p, dividend, length, sizeof(AM_LONG)); |
288
|
|
|
|
|
|
|
/* Magic number here... */ |
289
|
6243
|
|
|
|
|
|
outptr = outspace + 54; |
290
|
|
|
|
|
|
|
|
291
|
|
|
|
|
|
|
while (1) { |
292
|
20647
|
|
|
|
|
|
AM_LONG *temp, carry = 0; |
293
|
70591
|
100
|
|
|
|
|
while (length && (*(dividend + length - 1) == 0)) --length; |
|
|
100
|
|
|
|
|
|
294
|
20647
|
100
|
|
|
|
|
if (length == 0) { |
295
|
6243
|
|
|
|
|
|
sv_setpvn(s, outptr, outlength); |
296
|
6243
|
|
|
|
|
|
break; |
297
|
|
|
|
|
|
|
} |
298
|
14404
|
|
|
|
|
|
dptr = dividend + length - 1; |
299
|
14404
|
|
|
|
|
|
qptr = quotient + length - 1; |
300
|
28812
|
100
|
|
|
|
|
while (dptr >= dividend) { |
301
|
|
|
|
|
|
|
unsigned int i; |
302
|
14408
|
|
|
|
|
|
*dptr += carry << 16; |
303
|
14408
|
|
|
|
|
|
*qptr = 0; |
304
|
244936
|
100
|
|
|
|
|
for (i = 16; i; ) { |
305
|
230528
|
|
|
|
|
|
--i; |
306
|
230528
|
100
|
|
|
|
|
if (tens[i] <= *dptr) { |
307
|
17580
|
|
|
|
|
|
*dptr -= tens[i]; |
308
|
17580
|
|
|
|
|
|
*qptr += ones[i]; |
309
|
|
|
|
|
|
|
} |
310
|
|
|
|
|
|
|
} |
311
|
14408
|
|
|
|
|
|
carry = *dptr; |
312
|
14408
|
|
|
|
|
|
--dptr; |
313
|
14408
|
|
|
|
|
|
--qptr; |
314
|
|
|
|
|
|
|
} |
315
|
14404
|
|
|
|
|
|
--outptr; |
316
|
14404
|
|
|
|
|
|
*outptr = (char) (0x30 + *dividend) & 0x00ff; |
317
|
14404
|
|
|
|
|
|
++outlength; |
318
|
14404
|
|
|
|
|
|
temp = dividend; |
319
|
14404
|
|
|
|
|
|
dividend = quotient; |
320
|
14404
|
|
|
|
|
|
quotient = temp; |
321
|
14404
|
|
|
|
|
|
} |
322
|
|
|
|
|
|
|
|
323
|
6243
|
|
|
|
|
|
SvNVX(s) = nn; |
324
|
6243
|
|
|
|
|
|
SvNOK_on(s); |
325
|
6243
|
|
|
|
|
|
} |
326
|
|
|
|
|
|
|
|
327
|
|
|
|
|
|
|
/* Given 2 lists of training item indices sorted in descending order, |
328
|
|
|
|
|
|
|
* fill a third list with the intersection of items in these lists. |
329
|
|
|
|
|
|
|
* This is a simple intersection, and no check for heterogeneity is |
330
|
|
|
|
|
|
|
* performed. |
331
|
|
|
|
|
|
|
* Return the next empty (available) index address in the third list. |
332
|
|
|
|
|
|
|
* If the two lists have no intersection, then the return value is |
333
|
|
|
|
|
|
|
* just the same as the third input. |
334
|
|
|
|
|
|
|
*/ |
335
|
9227
|
|
|
|
|
|
unsigned short *intersect_supras( |
336
|
|
|
|
|
|
|
AM_SHORT *i, AM_SHORT *j, AM_SHORT *k){ |
337
|
|
|
|
|
|
|
AM_SHORT *temp; |
338
|
|
|
|
|
|
|
while (1) { |
339
|
343093
|
100
|
|
|
|
|
while (*i > *j) |
340
|
221153
|
|
|
|
|
|
--i; |
341
|
121940
|
100
|
|
|
|
|
if (*i == 0) break; |
342
|
112713
|
100
|
|
|
|
|
if (*i < *j) { |
343
|
33130
|
|
|
|
|
|
temp = i; |
344
|
33130
|
|
|
|
|
|
i = j; |
345
|
33130
|
|
|
|
|
|
j = temp; |
346
|
33130
|
|
|
|
|
|
continue; |
347
|
|
|
|
|
|
|
} |
348
|
79583
|
|
|
|
|
|
*k = *i; |
349
|
79583
|
|
|
|
|
|
--i; |
350
|
79583
|
|
|
|
|
|
--j; |
351
|
79583
|
|
|
|
|
|
--k; |
352
|
112713
|
|
|
|
|
|
} |
353
|
9227
|
|
|
|
|
|
return k; |
354
|
|
|
|
|
|
|
} |
355
|
|
|
|
|
|
|
/* The first three inputs are the same as for intersect_supra above, |
356
|
|
|
|
|
|
|
* and the fourth paramater should be a list containing the class |
357
|
|
|
|
|
|
|
* index for all of the training items. In addition to combining |
358
|
|
|
|
|
|
|
* the first two lists into the third via intersection, the final |
359
|
|
|
|
|
|
|
* list is checked for heterogeneity and the non-deterministic |
360
|
|
|
|
|
|
|
* heterogeneous supracontexts are removed. |
361
|
|
|
|
|
|
|
* The return value is the number of items contained in the resulting |
362
|
|
|
|
|
|
|
* list. |
363
|
|
|
|
|
|
|
*/ |
364
|
33728
|
|
|
|
|
|
AM_SHORT intersect_supras_final( |
365
|
|
|
|
|
|
|
AM_SHORT *i, AM_SHORT *j, |
366
|
|
|
|
|
|
|
AM_SHORT *intersect, AM_SHORT *subcontext_class){ |
367
|
33728
|
|
|
|
|
|
AM_SHORT class = 0; |
368
|
33728
|
|
|
|
|
|
AM_SHORT length = 0; |
369
|
|
|
|
|
|
|
AM_SHORT *temp; |
370
|
|
|
|
|
|
|
while (1) { |
371
|
699168
|
100
|
|
|
|
|
while (*i > *j) |
372
|
516500
|
|
|
|
|
|
--i; |
373
|
182668
|
100
|
|
|
|
|
if (*i == 0) |
374
|
21261
|
|
|
|
|
|
break; |
375
|
161407
|
100
|
|
|
|
|
if (*i < *j) { |
376
|
91372
|
|
|
|
|
|
temp = i; |
377
|
91372
|
|
|
|
|
|
i = j; |
378
|
91372
|
|
|
|
|
|
j = temp; |
379
|
91372
|
|
|
|
|
|
continue; |
380
|
|
|
|
|
|
|
} |
381
|
70035
|
|
|
|
|
|
*intersect = *i; |
382
|
70035
|
|
|
|
|
|
++intersect; |
383
|
70035
|
|
|
|
|
|
++length; |
384
|
|
|
|
|
|
|
|
385
|
|
|
|
|
|
|
/* is it heterogeneous? */ |
386
|
70035
|
100
|
|
|
|
|
if (class == 0) { |
387
|
|
|
|
|
|
|
/* is it not deterministic? */ |
388
|
29851
|
100
|
|
|
|
|
if (length > 1) { |
389
|
1763
|
|
|
|
|
|
length = 0; |
390
|
1763
|
|
|
|
|
|
break; |
391
|
|
|
|
|
|
|
} else { |
392
|
28088
|
|
|
|
|
|
class = subcontext_class[*i]; |
393
|
|
|
|
|
|
|
} |
394
|
|
|
|
|
|
|
} else { |
395
|
|
|
|
|
|
|
/* Do the classes not match? */ |
396
|
40184
|
100
|
|
|
|
|
if (class != subcontext_class[*i]) { |
397
|
10704
|
|
|
|
|
|
length = 0; |
398
|
10704
|
|
|
|
|
|
break; |
399
|
|
|
|
|
|
|
} |
400
|
|
|
|
|
|
|
} |
401
|
57568
|
|
|
|
|
|
--i; |
402
|
57568
|
|
|
|
|
|
--j; |
403
|
148940
|
|
|
|
|
|
} |
404
|
33728
|
|
|
|
|
|
return length; |
405
|
|
|
|
|
|
|
} |
406
|
|
|
|
|
|
|
|
407
|
|
|
|
|
|
|
MODULE = Algorithm::AM PACKAGE = Algorithm::AM |
408
|
|
|
|
|
|
|
|
409
|
|
|
|
|
|
|
BOOT: |
410
|
|
|
|
|
|
|
{ |
411
|
10
|
|
|
|
|
|
AM_LONG ten = 10; |
412
|
10
|
|
|
|
|
|
AM_LONG one = 1; |
413
|
10
|
|
|
|
|
|
AM_LONG *tensptr = &tens[0]; |
414
|
10
|
|
|
|
|
|
AM_LONG *onesptr = &ones[0]; |
415
|
|
|
|
|
|
|
unsigned int i; |
416
|
170
|
100
|
|
|
|
|
for (i = 16; i; i--) { |
417
|
160
|
|
|
|
|
|
*tensptr = ten; |
418
|
160
|
|
|
|
|
|
*onesptr = one; |
419
|
160
|
|
|
|
|
|
++tensptr; |
420
|
160
|
|
|
|
|
|
++onesptr; |
421
|
160
|
|
|
|
|
|
ten <<= 1; |
422
|
160
|
|
|
|
|
|
one <<= 1; |
423
|
|
|
|
|
|
|
} |
424
|
|
|
|
|
|
|
} |
425
|
|
|
|
|
|
|
|
426
|
|
|
|
|
|
|
/* |
427
|
|
|
|
|
|
|
* This function is called by from AM.pm right after creating |
428
|
|
|
|
|
|
|
* a blessed reference to Algorithm::AM. It stores the necessary |
429
|
|
|
|
|
|
|
* pointers in the AM_GUTS structure and attaches it to the magic |
430
|
|
|
|
|
|
|
* part of thre reference. |
431
|
|
|
|
|
|
|
* |
432
|
|
|
|
|
|
|
*/ |
433
|
|
|
|
|
|
|
|
434
|
|
|
|
|
|
|
void |
435
|
|
|
|
|
|
|
_xs_initialize(...) |
436
|
|
|
|
|
|
|
PREINIT: |
437
|
|
|
|
|
|
|
HV *project; |
438
|
|
|
|
|
|
|
AM_GUTS guts; /* NOT A POINTER THIS TIME! (let memory allocate automatically) */ |
439
|
|
|
|
|
|
|
SV **lattice_sizes; |
440
|
|
|
|
|
|
|
SV *svguts; |
441
|
|
|
|
|
|
|
MAGIC *mg; |
442
|
|
|
|
|
|
|
int i; |
443
|
|
|
|
|
|
|
PPCODE: |
444
|
|
|
|
|
|
|
/* 9 arguments are passed to the _xs_initialize method: */ |
445
|
|
|
|
|
|
|
/* $self, the AM object */ |
446
|
192
|
|
|
|
|
|
project = (HV *) SvRV(ST(0)); |
447
|
|
|
|
|
|
|
/* For explanations on these, see the comments on AM_guts */ |
448
|
192
|
|
|
|
|
|
lattice_sizes = AvARRAY((AV *) SvRV(ST(1))); |
449
|
192
|
|
|
|
|
|
guts.classes = AvARRAY((AV *) SvRV(ST(2))); |
450
|
192
|
|
|
|
|
|
guts.itemcontextchain = AvARRAY((AV *) SvRV(ST(3))); |
451
|
192
|
|
|
|
|
|
guts.itemcontextchainhead = (HV *) SvRV(ST(4)); |
452
|
192
|
|
|
|
|
|
guts.context_to_class = (HV *) SvRV(ST(5)); |
453
|
192
|
|
|
|
|
|
guts.contextsize = (HV *) SvRV(ST(6)); |
454
|
192
|
|
|
|
|
|
guts.pointers = (HV *) SvRV(ST(7)); |
455
|
192
|
|
|
|
|
|
guts.gang = (HV *) SvRV(ST(8)); |
456
|
192
|
|
|
|
|
|
guts.sum = AvARRAY((AV *) SvRV(ST(9))); |
457
|
192
|
|
|
|
|
|
guts.num_classes = av_len((AV *) SvRV(ST(9))); |
458
|
|
|
|
|
|
|
|
459
|
|
|
|
|
|
|
/* |
460
|
|
|
|
|
|
|
* Since the sublattices are small, we just take a chunk of memory |
461
|
|
|
|
|
|
|
* here that will be large enough for our purposes and do the actual |
462
|
|
|
|
|
|
|
* memory allocation within the code; this reduces the overhead of |
463
|
|
|
|
|
|
|
* repeated system calls. |
464
|
|
|
|
|
|
|
* |
465
|
|
|
|
|
|
|
*/ |
466
|
|
|
|
|
|
|
|
467
|
960
|
100
|
|
|
|
|
for (i = 0; i < NUM_LATTICES; ++i) { |
468
|
768
|
|
|
|
|
|
UV v = SvUVX(lattice_sizes[i]); |
469
|
768
|
50
|
|
|
|
|
Newz(0, guts.lptr[i], 1 << v, AM_SHORT); |
470
|
768
|
50
|
|
|
|
|
Newz(0, guts.sptr[i], 1 << (v + 1), AM_SUPRA); /* CHANGED */ |
471
|
768
|
|
|
|
|
|
Newz(0, guts.sptr[i][0].data, 2, AM_SHORT); |
472
|
|
|
|
|
|
|
} |
473
|
|
|
|
|
|
|
|
474
|
|
|
|
|
|
|
/* Perl magic invoked here */ |
475
|
|
|
|
|
|
|
|
476
|
192
|
|
|
|
|
|
svguts = newSVpv((char *) &guts, sizeof(AM_GUTS)); |
477
|
192
|
|
|
|
|
|
sv_magic((SV *) project, svguts, PERL_MAGIC_ext, NULL, 0); |
478
|
192
|
|
|
|
|
|
SvRMAGICAL_off((SV *) project); |
479
|
192
|
|
|
|
|
|
mg = mg_find((SV *) project, PERL_MAGIC_ext); |
480
|
192
|
|
|
|
|
|
mg->mg_virtual = &AMguts_vtab; |
481
|
192
|
|
|
|
|
|
mg_magical((SV *) project); |
482
|
|
|
|
|
|
|
|
483
|
|
|
|
|
|
|
void |
484
|
|
|
|
|
|
|
_fillandcount(...) |
485
|
|
|
|
|
|
|
PREINIT: |
486
|
|
|
|
|
|
|
HV *project; |
487
|
|
|
|
|
|
|
UV linear_flag; |
488
|
|
|
|
|
|
|
AM_GUTS *guts; |
489
|
|
|
|
|
|
|
MAGIC *mg; |
490
|
|
|
|
|
|
|
SV **lattice_sizes_input; |
491
|
|
|
|
|
|
|
AM_SHORT lattice_sizes[NUM_LATTICES]; |
492
|
|
|
|
|
|
|
AM_SHORT **lptr; |
493
|
|
|
|
|
|
|
AM_SUPRA **sptr; |
494
|
|
|
|
|
|
|
AM_SHORT nptr[NUM_LATTICES];/* this helps us manage the free list in sptr[i] */ |
495
|
|
|
|
|
|
|
AM_SHORT subcontextnumber; |
496
|
|
|
|
|
|
|
AM_SHORT *subcontext; |
497
|
|
|
|
|
|
|
AM_SHORT *subcontext_class; |
498
|
|
|
|
|
|
|
SV **classes, **itemcontextchain, **sum; |
499
|
|
|
|
|
|
|
HV *itemcontextchainhead, *context_to_class, *contextsize, *pointers, *gang; |
500
|
|
|
|
|
|
|
IV num_classes; |
501
|
|
|
|
|
|
|
HE *he; |
502
|
194
|
|
|
|
|
|
AM_BIG_INT grandtotal = {0, 0, 0, 0, 0, 0, 0, 0}; |
503
|
|
|
|
|
|
|
SV *tempsv; |
504
|
|
|
|
|
|
|
int chunk, i; |
505
|
|
|
|
|
|
|
AM_SHORT gaps[16]; |
506
|
|
|
|
|
|
|
AM_SHORT *intersect, *intersectlist; |
507
|
|
|
|
|
|
|
AM_SHORT *intersectlist2, *intersectlist3, *ilist2top, *ilist3top; |
508
|
|
|
|
|
|
|
PPCODE: |
509
|
|
|
|
|
|
|
/* Input args are the AM object ($self), number of features |
510
|
|
|
|
|
|
|
* perl lattice, and a flag to indicate whether to count occurrences |
511
|
|
|
|
|
|
|
* (true) or pointers (false), also known as linear/quadratic. |
512
|
|
|
|
|
|
|
*/ |
513
|
194
|
|
|
|
|
|
project = (HV *) SvRV(ST(0)); |
514
|
194
|
|
|
|
|
|
lattice_sizes_input = AvARRAY((AV *) SvRV(ST(1))); |
515
|
194
|
|
|
|
|
|
linear_flag = SvUVX(ST(2)); |
516
|
194
|
|
|
|
|
|
mg = mg_find((SV *) project, PERL_MAGIC_ext); |
517
|
194
|
|
|
|
|
|
guts = (AM_GUTS *) SvPVX(mg->mg_obj); |
518
|
|
|
|
|
|
|
|
519
|
|
|
|
|
|
|
/* |
520
|
|
|
|
|
|
|
* We initialize the memory for the sublattices, including setting up the |
521
|
|
|
|
|
|
|
* linked lists. |
522
|
|
|
|
|
|
|
* |
523
|
|
|
|
|
|
|
*/ |
524
|
|
|
|
|
|
|
|
525
|
194
|
|
|
|
|
|
lptr = guts->lptr; |
526
|
194
|
|
|
|
|
|
sptr = guts->sptr; |
527
|
970
|
100
|
|
|
|
|
for (chunk = 0; chunk < NUM_LATTICES; ++chunk) { |
528
|
|
|
|
|
|
|
/* Extract numeric values for the specified lattice_sizes */ |
529
|
776
|
|
|
|
|
|
lattice_sizes[chunk] = (AM_SHORT) SvUVX(lattice_sizes_input[chunk]); |
530
|
|
|
|
|
|
|
/* TODO: explain the lines below */ |
531
|
776
|
50
|
|
|
|
|
Zero(lptr[chunk], 1 << lattice_sizes[chunk], AM_SHORT); |
532
|
776
|
|
|
|
|
|
sptr[chunk][0].next = 0; |
533
|
776
|
|
|
|
|
|
nptr[chunk] = 1; |
534
|
7048
|
100
|
|
|
|
|
for (i = 1; i < 1 << (lattice_sizes[chunk] + 1); ++i) /* CHANGED */ |
535
|
6272
|
|
|
|
|
|
sptr[chunk][i].next = (AM_SHORT) i + 1; |
536
|
|
|
|
|
|
|
} |
537
|
|
|
|
|
|
|
|
538
|
|
|
|
|
|
|
/* |
539
|
|
|
|
|
|
|
* Instead of adding subcontext labels directly to the supracontexts, |
540
|
|
|
|
|
|
|
* we store all of these labels in an array called subcontext. We |
541
|
|
|
|
|
|
|
* then store the array indices of the subcontext labels in the |
542
|
|
|
|
|
|
|
* supracontexts. That means the list of subcontexts in the |
543
|
|
|
|
|
|
|
* supracontexts is an increasing sequence of positive integers, handy |
544
|
|
|
|
|
|
|
* for taking intersections (see lattice.pod). |
545
|
|
|
|
|
|
|
* |
546
|
|
|
|
|
|
|
* The index into the array is called subcontextnumber. |
547
|
|
|
|
|
|
|
* |
548
|
|
|
|
|
|
|
* The array of matching classes is called subcontext_class. |
549
|
|
|
|
|
|
|
* |
550
|
|
|
|
|
|
|
*/ |
551
|
|
|
|
|
|
|
|
552
|
194
|
|
|
|
|
|
context_to_class = guts->context_to_class; |
553
|
194
|
50
|
|
|
|
|
subcontextnumber = (AM_SHORT) HvUSEDKEYS(context_to_class); |
554
|
194
|
50
|
|
|
|
|
Newz(0, subcontext, NUM_LATTICES * (subcontextnumber + 1), AM_SHORT); |
555
|
194
|
|
|
|
|
|
subcontext += NUM_LATTICES * subcontextnumber; |
556
|
194
|
50
|
|
|
|
|
Newz(0, subcontext_class, subcontextnumber + 1, AM_SHORT); |
557
|
194
|
|
|
|
|
|
subcontext_class += subcontextnumber; |
558
|
194
|
50
|
|
|
|
|
Newz(0, intersectlist, subcontextnumber + 1, AM_SHORT); |
559
|
194
|
50
|
|
|
|
|
Newz(0, intersectlist2, subcontextnumber + 1, AM_SHORT); |
560
|
194
|
|
|
|
|
|
ilist2top = intersectlist2 + subcontextnumber; |
561
|
194
|
50
|
|
|
|
|
Newz(0, intersectlist3, subcontextnumber + 1, AM_SHORT); |
562
|
194
|
|
|
|
|
|
ilist3top = intersectlist3 + subcontextnumber; |
563
|
|
|
|
|
|
|
|
564
|
194
|
|
|
|
|
|
hv_iterinit(context_to_class); |
565
|
8382
|
100
|
|
|
|
|
while (he = hv_iternext(context_to_class)) { |
566
|
8188
|
|
|
|
|
|
AM_SHORT *contextptr = (AM_SHORT *) HeKEY(he); |
567
|
8188
|
|
|
|
|
|
AM_SHORT class = (AM_SHORT) SvUVX(HeVAL(he)); |
568
|
40940
|
100
|
|
|
|
|
for (chunk = 0; chunk < NUM_LATTICES; ++chunk, ++contextptr) { |
569
|
32752
|
|
|
|
|
|
AM_SHORT active = lattice_sizes[chunk]; |
570
|
32752
|
|
|
|
|
|
AM_SHORT *lattice = lptr[chunk]; |
571
|
32752
|
|
|
|
|
|
AM_SUPRA *supralist = sptr[chunk]; |
572
|
32752
|
|
|
|
|
|
AM_SHORT nextsupra = nptr[chunk]; |
573
|
32752
|
|
|
|
|
|
AM_SHORT context = *contextptr; |
574
|
|
|
|
|
|
|
AM_SUPRA *p, *c; |
575
|
|
|
|
|
|
|
AM_SHORT pi, ci; |
576
|
32752
|
|
|
|
|
|
AM_SHORT d, t, tt, numgaps = 0; |
577
|
|
|
|
|
|
|
|
578
|
|
|
|
|
|
|
/* We want to add subcontextnumber to the appropriate |
579
|
|
|
|
|
|
|
* supracontexts in the four smaller lattices. |
580
|
|
|
|
|
|
|
* |
581
|
|
|
|
|
|
|
* Suppose we want to add subcontextnumber to the supracontext |
582
|
|
|
|
|
|
|
* labeled by d. supralist[lattice[d]] is an AM_SUPRA which |
583
|
|
|
|
|
|
|
* reflects the current state of the supracontext. Suppose this |
584
|
|
|
|
|
|
|
* state is |
585
|
|
|
|
|
|
|
* |
586
|
|
|
|
|
|
|
* data: 2 0 x y (i.e., currently contains two subcontexts) |
587
|
|
|
|
|
|
|
* count: 5 |
588
|
|
|
|
|
|
|
* next: 7 |
589
|
|
|
|
|
|
|
* touched: 0 |
590
|
|
|
|
|
|
|
* |
591
|
|
|
|
|
|
|
* Then we pluck an unused AM_SUPRA off of the free list; |
592
|
|
|
|
|
|
|
* suppose that it's located at supralist[9] (the variable |
593
|
|
|
|
|
|
|
* nextsupra tells us where). Then supralist[lattice[d]] will |
594
|
|
|
|
|
|
|
* change to |
595
|
|
|
|
|
|
|
* |
596
|
|
|
|
|
|
|
* data: 2 0 x y |
597
|
|
|
|
|
|
|
* count: 4 (decrease by 1) |
598
|
|
|
|
|
|
|
* next: 9 |
599
|
|
|
|
|
|
|
* touched: 1 |
600
|
|
|
|
|
|
|
* |
601
|
|
|
|
|
|
|
* and supralist[9] will become |
602
|
|
|
|
|
|
|
* |
603
|
|
|
|
|
|
|
* data: 3 0 subcontextnumber x y (now has three subcontexts) |
604
|
|
|
|
|
|
|
* count: 1 |
605
|
|
|
|
|
|
|
* next: 7 |
606
|
|
|
|
|
|
|
* touched: 0 |
607
|
|
|
|
|
|
|
* |
608
|
|
|
|
|
|
|
* (note: the entries in data[] are added in decreasing order) |
609
|
|
|
|
|
|
|
* |
610
|
|
|
|
|
|
|
* |
611
|
|
|
|
|
|
|
* If, on the other hand, if supralist[lattice[d]] looks like |
612
|
|
|
|
|
|
|
* |
613
|
|
|
|
|
|
|
* data: 2 0 x y |
614
|
|
|
|
|
|
|
* count: 8 |
615
|
|
|
|
|
|
|
* next: 11 |
616
|
|
|
|
|
|
|
* touched: 1 |
617
|
|
|
|
|
|
|
* |
618
|
|
|
|
|
|
|
* that means that supralist[11] must look something like |
619
|
|
|
|
|
|
|
* |
620
|
|
|
|
|
|
|
* data: 3 0 subcontextnumber x y |
621
|
|
|
|
|
|
|
* count: 4 |
622
|
|
|
|
|
|
|
* next: 2 |
623
|
|
|
|
|
|
|
* touched: 0 |
624
|
|
|
|
|
|
|
* |
625
|
|
|
|
|
|
|
* There already exists a supracontext with subcontextnumber |
626
|
|
|
|
|
|
|
* added in! So we change supralist[lattice[d]] to |
627
|
|
|
|
|
|
|
* |
628
|
|
|
|
|
|
|
* data: 2 0 x y |
629
|
|
|
|
|
|
|
* count: 7 (decrease by 1) |
630
|
|
|
|
|
|
|
* next: 11 |
631
|
|
|
|
|
|
|
* touched: 1 |
632
|
|
|
|
|
|
|
* |
633
|
|
|
|
|
|
|
* change supralist[11] to |
634
|
|
|
|
|
|
|
* |
635
|
|
|
|
|
|
|
* data: 3 0 subcontextnumber x y |
636
|
|
|
|
|
|
|
* count: 5 (increase by 1) |
637
|
|
|
|
|
|
|
* next: 2 |
638
|
|
|
|
|
|
|
* touched: 0 |
639
|
|
|
|
|
|
|
* |
640
|
|
|
|
|
|
|
* and set lattice[d] = 11. |
641
|
|
|
|
|
|
|
*/ |
642
|
|
|
|
|
|
|
|
643
|
32752
|
|
|
|
|
|
subcontext[chunk] = context; |
644
|
|
|
|
|
|
|
|
645
|
32752
|
100
|
|
|
|
|
if (context == 0) { |
646
|
28745
|
100
|
|
|
|
|
for (p = supralist + supralist->next; |
647
|
22381
|
|
|
|
|
|
p != supralist; p = supralist + p->next) { |
648
|
|
|
|
|
|
|
AM_SHORT *data; |
649
|
22381
|
50
|
|
|
|
|
Newz(0, data, p->data[0] + 3, AM_SHORT); |
650
|
22381
|
|
|
|
|
|
Copy(p->data + 2, data + 3, p->data[0], AM_SHORT); |
651
|
22381
|
|
|
|
|
|
data[2] = subcontextnumber; |
652
|
22381
|
|
|
|
|
|
data[0] = p->data[0] + 1; |
653
|
22381
|
|
|
|
|
|
Safefree(p->data); |
654
|
22381
|
|
|
|
|
|
p->data = data; |
655
|
|
|
|
|
|
|
} |
656
|
6364
|
100
|
|
|
|
|
if (lattice[context] == 0) { |
657
|
|
|
|
|
|
|
|
658
|
|
|
|
|
|
|
/* in this case, the subcontext will be |
659
|
|
|
|
|
|
|
* added to all supracontexts, so there's |
660
|
|
|
|
|
|
|
* no need to hassle with a Gray code and |
661
|
|
|
|
|
|
|
* move pointers |
662
|
|
|
|
|
|
|
*/ |
663
|
|
|
|
|
|
|
|
664
|
732
|
|
|
|
|
|
AM_SHORT count = 0; |
665
|
732
|
|
|
|
|
|
ci = nptr[chunk]; |
666
|
732
|
|
|
|
|
|
nptr[chunk] = supralist[ci].next; |
667
|
732
|
|
|
|
|
|
c = supralist + ci; |
668
|
732
|
|
|
|
|
|
c->next = supralist->next; |
669
|
732
|
|
|
|
|
|
supralist->next = ci; |
670
|
732
|
|
|
|
|
|
Newz(0, c->data, 3, AM_SHORT); |
671
|
732
|
|
|
|
|
|
c->data[2] = subcontextnumber; |
672
|
732
|
|
|
|
|
|
c->data[0] = 1; |
673
|
3723
|
100
|
|
|
|
|
for (i = 0; i < (1 << active); ++i) { |
674
|
2991
|
100
|
|
|
|
|
if (lattice[i] == 0) { |
675
|
1616
|
|
|
|
|
|
lattice[i] = ci; |
676
|
1616
|
|
|
|
|
|
++count; |
677
|
|
|
|
|
|
|
} |
678
|
|
|
|
|
|
|
} |
679
|
732
|
|
|
|
|
|
c->count = count; |
680
|
|
|
|
|
|
|
} |
681
|
6364
|
|
|
|
|
|
continue; |
682
|
|
|
|
|
|
|
} |
683
|
|
|
|
|
|
|
|
684
|
|
|
|
|
|
|
/* set up traversal using Gray code */ |
685
|
26388
|
|
|
|
|
|
d = context; |
686
|
81785
|
100
|
|
|
|
|
for (i = 1 << (active - 1); i; i >>= 1) |
687
|
55397
|
100
|
|
|
|
|
if (!(i & context)) |
688
|
15490
|
|
|
|
|
|
gaps[numgaps++] = i; |
689
|
26388
|
|
|
|
|
|
t = 1 << numgaps; |
690
|
|
|
|
|
|
|
|
691
|
26388
|
|
|
|
|
|
p = supralist + (pi = lattice[context]); |
692
|
26388
|
100
|
|
|
|
|
if (pi) |
693
|
25342
|
|
|
|
|
|
--(p->count); |
694
|
26388
|
|
|
|
|
|
ci = nextsupra; |
695
|
26388
|
|
|
|
|
|
nextsupra = supralist[ci].next; |
696
|
26388
|
|
|
|
|
|
p->touched = 1; |
697
|
26388
|
|
|
|
|
|
c = supralist + ci; |
698
|
26388
|
|
|
|
|
|
c->touched = 0; |
699
|
26388
|
|
|
|
|
|
c->next = p->next; |
700
|
26388
|
|
|
|
|
|
p->next = ci; |
701
|
26388
|
|
|
|
|
|
c->count = 1; |
702
|
26388
|
50
|
|
|
|
|
Newz(0, c->data, p->data[0] + 3, AM_SHORT); |
703
|
26388
|
|
|
|
|
|
Copy(p->data + 2, c->data + 3, p->data[0], AM_SHORT); |
704
|
26388
|
|
|
|
|
|
c->data[2] = subcontextnumber; |
705
|
26388
|
|
|
|
|
|
c->data[0] = p->data[0] + 1; |
706
|
26388
|
|
|
|
|
|
lattice[context] = ci; |
707
|
|
|
|
|
|
|
|
708
|
|
|
|
|
|
|
/* traverse */ |
709
|
43317
|
100
|
|
|
|
|
while (--t) { |
710
|
|
|
|
|
|
|
/* find the rightmost 1 in t; from HAKMEM, I believe */ |
711
|
18368
|
100
|
|
|
|
|
for (i = 0, tt = ~t & (t - 1); tt; tt >>= 1, ++i) |
712
|
|
|
|
|
|
|
; |
713
|
16929
|
|
|
|
|
|
d ^= gaps[i]; |
714
|
|
|
|
|
|
|
|
715
|
16929
|
|
|
|
|
|
p = supralist + (pi = lattice[d]); |
716
|
16929
|
100
|
|
|
|
|
if (pi) |
717
|
16404
|
|
|
|
|
|
--(p->count); |
718
|
16929
|
|
|
|
|
|
switch (p->touched) { |
719
|
|
|
|
|
|
|
case 1: |
720
|
1158
|
|
|
|
|
|
++supralist[lattice[d] = p->next].count; |
721
|
1158
|
|
|
|
|
|
break; |
722
|
|
|
|
|
|
|
case 0: |
723
|
15771
|
|
|
|
|
|
ci = nextsupra; |
724
|
15771
|
|
|
|
|
|
nextsupra = supralist[ci].next; |
725
|
15771
|
|
|
|
|
|
p->touched = 1; |
726
|
15771
|
|
|
|
|
|
c = supralist + ci; |
727
|
15771
|
|
|
|
|
|
c->touched = 0; |
728
|
15771
|
|
|
|
|
|
c->next = p->next; |
729
|
15771
|
|
|
|
|
|
p->next = ci; |
730
|
15771
|
|
|
|
|
|
c->count = 1; |
731
|
15771
|
50
|
|
|
|
|
Newz(0, c->data, p->data[0] + 3, AM_SHORT); |
732
|
15771
|
|
|
|
|
|
Copy(p->data + 2, c->data + 3, p->data[0], AM_SHORT); |
733
|
15771
|
|
|
|
|
|
c->data[2] = subcontextnumber; |
734
|
15771
|
|
|
|
|
|
c->data[0] = p->data[0] + 1; |
735
|
15771
|
|
|
|
|
|
lattice[d] = ci; |
736
|
|
|
|
|
|
|
} |
737
|
|
|
|
|
|
|
} |
738
|
|
|
|
|
|
|
|
739
|
|
|
|
|
|
|
/* Here we return all AM_SUPRA with count 0 back to the free |
740
|
|
|
|
|
|
|
* list and set touched = 0 for all remaining. |
741
|
|
|
|
|
|
|
*/ |
742
|
|
|
|
|
|
|
|
743
|
26388
|
|
|
|
|
|
p = supralist; |
744
|
26388
|
|
|
|
|
|
p->touched = 0; |
745
|
|
|
|
|
|
|
do { |
746
|
136061
|
100
|
|
|
|
|
if (supralist[i = p->next].count == 0) { |
747
|
39980
|
|
|
|
|
|
Safefree(supralist[i].data); |
748
|
39980
|
|
|
|
|
|
p->next = supralist[i].next; |
749
|
39980
|
|
|
|
|
|
supralist[i].next = nextsupra; |
750
|
39980
|
|
|
|
|
|
nextsupra = (AM_SHORT) i; |
751
|
|
|
|
|
|
|
} else { |
752
|
96081
|
|
|
|
|
|
p = supralist + p->next; |
753
|
96081
|
|
|
|
|
|
p->touched = 0; |
754
|
|
|
|
|
|
|
} |
755
|
136061
|
100
|
|
|
|
|
} while (p->next); |
756
|
26388
|
|
|
|
|
|
nptr[chunk] = nextsupra; |
757
|
|
|
|
|
|
|
}/*end for(chunk = 0...*/ |
758
|
8188
|
|
|
|
|
|
subcontext -= NUM_LATTICES; |
759
|
8188
|
|
|
|
|
|
*subcontext_class = class; |
760
|
8188
|
|
|
|
|
|
--subcontext_class; |
761
|
8188
|
|
|
|
|
|
--subcontextnumber; |
762
|
|
|
|
|
|
|
}/*end while (he = hv_iternext(...*/ |
763
|
|
|
|
|
|
|
|
764
|
194
|
|
|
|
|
|
contextsize = guts->contextsize; |
765
|
194
|
|
|
|
|
|
pointers = guts->pointers; |
766
|
|
|
|
|
|
|
|
767
|
|
|
|
|
|
|
/* |
768
|
|
|
|
|
|
|
* The code is in three parts: |
769
|
|
|
|
|
|
|
* |
770
|
|
|
|
|
|
|
* 1. We successively take one nonempty supracontext from each of the |
771
|
|
|
|
|
|
|
* four small lattices and take their intersection to find a |
772
|
|
|
|
|
|
|
* supracontext of the big lattice. If at any point we get the |
773
|
|
|
|
|
|
|
* empty set, we move on. |
774
|
|
|
|
|
|
|
* |
775
|
|
|
|
|
|
|
* 2. We determine if the supracontext so found is heterogeneous; if |
776
|
|
|
|
|
|
|
* so, we skip it. |
777
|
|
|
|
|
|
|
* |
778
|
|
|
|
|
|
|
* 3. Otherwise, we count the pointers or occurrences. |
779
|
|
|
|
|
|
|
* |
780
|
|
|
|
|
|
|
*/ |
781
|
|
|
|
|
|
|
{ |
782
|
|
|
|
|
|
|
AM_SUPRA *p0, *p1, *p2, *p3; |
783
|
|
|
|
|
|
|
AM_SHORT length; |
784
|
|
|
|
|
|
|
AM_SHORT *k; |
785
|
|
|
|
|
|
|
|
786
|
|
|
|
|
|
|
/* find intersections */ |
787
|
824
|
100
|
|
|
|
|
for (p0 = sptr[0] + sptr[0]->next; p0 != sptr[0]; p0 = sptr[0] + p0->next) { |
788
|
2751
|
100
|
|
|
|
|
for (p1 = sptr[1] + sptr[1]->next; p1 != sptr[1]; p1 = sptr[1] + p1->next) { |
789
|
|
|
|
|
|
|
/*Find intersection between p0 and p2*/ |
790
|
2121
|
|
|
|
|
|
k = intersect_supras( |
791
|
2121
|
|
|
|
|
|
p0->data + p0->data[0] + 1, |
792
|
2121
|
|
|
|
|
|
p1->data + p1->data[0] + 1, |
793
|
|
|
|
|
|
|
ilist2top |
794
|
|
|
|
|
|
|
); |
795
|
|
|
|
|
|
|
/* If k has not been increased then intersection was empty */ |
796
|
2121
|
100
|
|
|
|
|
if (k == ilist2top) |
797
|
154
|
|
|
|
|
|
continue; |
798
|
1967
|
|
|
|
|
|
*k = 0; |
799
|
|
|
|
|
|
|
|
800
|
9073
|
100
|
|
|
|
|
for (p2 = sptr[2] + sptr[2]->next; p2 != sptr[2]; p2 = sptr[2] + p2->next) { |
801
|
|
|
|
|
|
|
|
802
|
|
|
|
|
|
|
/*Find intersection between previous intersection and p2*/ |
803
|
7106
|
|
|
|
|
|
k = intersect_supras( |
804
|
|
|
|
|
|
|
ilist2top, |
805
|
7106
|
|
|
|
|
|
p2->data + p2->data[0] + 1, |
806
|
|
|
|
|
|
|
ilist3top |
807
|
|
|
|
|
|
|
); |
808
|
|
|
|
|
|
|
/* If k has not been increased then intersection was empty */ |
809
|
7106
|
100
|
|
|
|
|
if (k == ilist3top) |
810
|
694
|
|
|
|
|
|
continue; |
811
|
6412
|
|
|
|
|
|
*k = 0; |
812
|
|
|
|
|
|
|
|
813
|
40140
|
100
|
|
|
|
|
for (p3 = sptr[3] + sptr[3]->next; p3 != sptr[3]; p3 = sptr[3] + p3->next) { |
814
|
|
|
|
|
|
|
|
815
|
|
|
|
|
|
|
/* Find intersection between previous intersection and p3; |
816
|
|
|
|
|
|
|
* check for disqualified supras this time. |
817
|
|
|
|
|
|
|
*/ |
818
|
33728
|
|
|
|
|
|
length = intersect_supras_final( |
819
|
|
|
|
|
|
|
ilist3top, |
820
|
33728
|
|
|
|
|
|
p3->data + p3->data[0] + 1, |
821
|
|
|
|
|
|
|
intersectlist, |
822
|
|
|
|
|
|
|
subcontext_class |
823
|
|
|
|
|
|
|
); |
824
|
|
|
|
|
|
|
|
825
|
|
|
|
|
|
|
/* count occurrences */ |
826
|
33728
|
100
|
|
|
|
|
if (length) { |
827
|
|
|
|
|
|
|
AM_SHORT i; |
828
|
15621
|
|
|
|
|
|
AM_BIG_INT count = {0, 0, 0, 0, 0, 0, 0, 0}; |
829
|
15621
|
|
|
|
|
|
AM_LONG mask = 0xffff; |
830
|
|
|
|
|
|
|
|
831
|
15621
|
|
|
|
|
|
count[0] = p0->count; |
832
|
|
|
|
|
|
|
|
833
|
15621
|
|
|
|
|
|
count[0] *= p1->count; |
834
|
15621
|
|
|
|
|
|
carry(count, 0); |
835
|
|
|
|
|
|
|
|
836
|
15621
|
|
|
|
|
|
count[0] *= p2->count; |
837
|
15621
|
|
|
|
|
|
count[1] *= p2->count; |
838
|
15621
|
|
|
|
|
|
carry(count, 0); |
839
|
15621
|
|
|
|
|
|
carry(count, 1); |
840
|
|
|
|
|
|
|
|
841
|
15621
|
|
|
|
|
|
count[0] *= p3->count; |
842
|
15621
|
|
|
|
|
|
count[1] *= p3->count; |
843
|
15621
|
|
|
|
|
|
count[2] *= p3->count; |
844
|
15621
|
|
|
|
|
|
carry(count, 0); |
845
|
15621
|
|
|
|
|
|
carry(count, 1); |
846
|
15621
|
|
|
|
|
|
carry(count, 2); |
847
|
15621
|
100
|
|
|
|
|
if(!linear_flag){ |
848
|
|
|
|
|
|
|
/* If scoring is pointers (quadratic) instead of linear*/ |
849
|
15601
|
|
|
|
|
|
AM_LONG pointercount = 0; |
850
|
50558
|
100
|
|
|
|
|
for (i = 0; i < length; ++i) |
851
|
34957
|
50
|
|
|
|
|
pointercount += (AM_LONG) SvUV(*hv_fetch(contextsize, |
852
|
|
|
|
|
|
|
(char *) (subcontext + (NUM_LATTICES * intersectlist[i])), 8, 0)); |
853
|
15601
|
50
|
|
|
|
|
if (pointercount & 0xffff0000) { |
854
|
0
|
|
|
|
|
|
AM_SHORT pchi = (AM_SHORT) (high_bits(pointercount)); |
855
|
0
|
|
|
|
|
|
AM_SHORT pclo = (AM_SHORT) (low_bits(pointercount)); |
856
|
|
|
|
|
|
|
AM_LONG hiprod[6]; |
857
|
0
|
|
|
|
|
|
hiprod[1] = pchi * count[0]; |
858
|
0
|
|
|
|
|
|
hiprod[2] = pchi * count[1]; |
859
|
0
|
|
|
|
|
|
hiprod[3] = pchi * count[2]; |
860
|
0
|
|
|
|
|
|
hiprod[4] = pchi * count[3]; |
861
|
0
|
|
|
|
|
|
count[0] *= pclo; |
862
|
0
|
|
|
|
|
|
count[1] *= pclo; |
863
|
0
|
|
|
|
|
|
count[2] *= pclo; |
864
|
0
|
|
|
|
|
|
count[3] *= pclo; |
865
|
0
|
|
|
|
|
|
carry(count, 0); |
866
|
0
|
|
|
|
|
|
carry(count, 1); |
867
|
0
|
|
|
|
|
|
carry(count, 2); |
868
|
0
|
|
|
|
|
|
carry(count, 3); |
869
|
|
|
|
|
|
|
|
870
|
0
|
|
|
|
|
|
count[1] += hiprod[1]; |
871
|
0
|
|
|
|
|
|
count[2] += hiprod[2]; |
872
|
0
|
|
|
|
|
|
count[3] += hiprod[3]; |
873
|
0
|
|
|
|
|
|
count[4] += hiprod[4]; |
874
|
0
|
|
|
|
|
|
carry(count, 1); |
875
|
0
|
|
|
|
|
|
carry(count, 2); |
876
|
0
|
|
|
|
|
|
carry(count, 3); |
877
|
0
|
|
|
|
|
|
carry(count, 4); |
878
|
|
|
|
|
|
|
} else { |
879
|
15601
|
|
|
|
|
|
count[0] *= pointercount; |
880
|
15601
|
|
|
|
|
|
count[1] *= pointercount; |
881
|
15601
|
|
|
|
|
|
count[2] *= pointercount; |
882
|
15601
|
|
|
|
|
|
count[3] *= pointercount; |
883
|
15601
|
|
|
|
|
|
carry(count, 0); |
884
|
15601
|
|
|
|
|
|
carry(count, 1); |
885
|
15601
|
|
|
|
|
|
carry(count, 2); |
886
|
15601
|
|
|
|
|
|
carry(count, 3); |
887
|
|
|
|
|
|
|
} |
888
|
|
|
|
|
|
|
} |
889
|
50611
|
100
|
|
|
|
|
for (i = 0; i < length; ++i) { |
890
|
|
|
|
|
|
|
int j; |
891
|
|
|
|
|
|
|
SV *tempsv; |
892
|
|
|
|
|
|
|
AM_LONG *p; |
893
|
34990
|
|
|
|
|
|
tempsv = *hv_fetch(pointers, |
894
|
|
|
|
|
|
|
(char *) (subcontext + (NUM_LATTICES * intersectlist[i])), 8, 1); |
895
|
34990
|
100
|
|
|
|
|
if (!SvPOK(tempsv)) { |
896
|
2745
|
50
|
|
|
|
|
SvUPGRADE(tempsv, SVt_PVNV); |
897
|
2745
|
50
|
|
|
|
|
SvGROW(tempsv, 8 * sizeof(AM_LONG) + 1); |
|
|
50
|
|
|
|
|
|
898
|
2745
|
|
|
|
|
|
Zero(SvPVX(tempsv), 8, AM_LONG); |
899
|
2745
|
|
|
|
|
|
SvCUR_set(tempsv, 8 * sizeof(AM_LONG)); |
900
|
2745
|
|
|
|
|
|
SvPOK_on(tempsv); |
901
|
|
|
|
|
|
|
} |
902
|
34990
|
|
|
|
|
|
p = (AM_LONG *) SvPVX(tempsv); |
903
|
279920
|
100
|
|
|
|
|
for (j = 0; j < 7; ++j) { |
904
|
244930
|
|
|
|
|
|
*(p + j) += count[j]; |
905
|
244930
|
|
|
|
|
|
carry_pointer(p + j); |
906
|
|
|
|
|
|
|
} |
907
|
|
|
|
|
|
|
}/* end for (i = 0;... */ |
908
|
|
|
|
|
|
|
}/* end if (length) */ |
909
|
|
|
|
|
|
|
}/* end for (p3 = sptr[3]... */ |
910
|
|
|
|
|
|
|
}/* end for (p2 = sptr[2]... */ |
911
|
|
|
|
|
|
|
}/* end for (p1 = sptr[1]... */ |
912
|
|
|
|
|
|
|
}/* end for (p0 = sptr[0]... */ |
913
|
|
|
|
|
|
|
/* clear out the supracontexts */ |
914
|
824
|
100
|
|
|
|
|
for (p0 = sptr[0] + sptr[0]->next; p0 != sptr[0]; p0 = sptr[0] + p0->next) |
915
|
630
|
|
|
|
|
|
Safefree(p0->data); |
916
|
833
|
100
|
|
|
|
|
for (p1 = sptr[1] + sptr[1]->next; p1 != sptr[1]; p1 = sptr[1] + p1->next) |
917
|
639
|
|
|
|
|
|
Safefree(p1->data); |
918
|
876
|
100
|
|
|
|
|
for (p2 = sptr[2] + sptr[2]->next; p2 != sptr[2]; p2 = sptr[2] + p2->next) |
919
|
682
|
|
|
|
|
|
Safefree(p2->data); |
920
|
1154
|
100
|
|
|
|
|
for (p3 = sptr[3] + sptr[3]->next; p3 != sptr[3]; p3 = sptr[3] + p3->next) |
921
|
960
|
|
|
|
|
|
Safefree(p3->data); |
922
|
|
|
|
|
|
|
|
923
|
|
|
|
|
|
|
/* |
924
|
|
|
|
|
|
|
* compute analogical set and gang effects |
925
|
|
|
|
|
|
|
* |
926
|
|
|
|
|
|
|
* Technically, we don't compute the analogical set; instead, we |
927
|
|
|
|
|
|
|
* compute how many pointers/occurrences there are for each of the |
928
|
|
|
|
|
|
|
* data items in a particular subcontext, and associate that number |
929
|
|
|
|
|
|
|
* with the subcontext label, not directly with the data item. We can |
930
|
|
|
|
|
|
|
* do this because if two data items are in the same subcontext, they |
931
|
|
|
|
|
|
|
* will have the same number of pointers/occurrences. |
932
|
|
|
|
|
|
|
* |
933
|
|
|
|
|
|
|
* If the user wants the detailed analogical set, it will be created |
934
|
|
|
|
|
|
|
* in Result.pm. |
935
|
|
|
|
|
|
|
* |
936
|
|
|
|
|
|
|
*/ |
937
|
|
|
|
|
|
|
|
938
|
194
|
|
|
|
|
|
gang = guts->gang; |
939
|
194
|
|
|
|
|
|
classes = guts->classes; |
940
|
194
|
|
|
|
|
|
itemcontextchain = guts->itemcontextchain; |
941
|
194
|
|
|
|
|
|
itemcontextchainhead = guts->itemcontextchainhead; |
942
|
194
|
|
|
|
|
|
sum = guts->sum; |
943
|
194
|
|
|
|
|
|
num_classes = guts->num_classes; |
944
|
194
|
|
|
|
|
|
hv_iterinit(pointers); |
945
|
2939
|
100
|
|
|
|
|
while (he = hv_iternext(pointers)) { |
946
|
|
|
|
|
|
|
AM_LONG count; |
947
|
|
|
|
|
|
|
AM_SHORT counthi, countlo; |
948
|
|
|
|
|
|
|
AM_BIG_INT p; |
949
|
|
|
|
|
|
|
AM_BIG_INT gangcount; |
950
|
|
|
|
|
|
|
AM_SHORT this_class; |
951
|
|
|
|
|
|
|
SV *dataitem; |
952
|
2745
|
|
|
|
|
|
Copy(SvPVX(HeVAL(he)), p, 8, AM_LONG); |
953
|
|
|
|
|
|
|
|
954
|
2745
|
|
|
|
|
|
tempsv = *hv_fetch(contextsize, HeKEY(he), NUM_LATTICES * sizeof(AM_SHORT), 0); |
955
|
2745
|
|
|
|
|
|
count = (AM_LONG) SvUVX(tempsv); |
956
|
2745
|
|
|
|
|
|
counthi = (AM_SHORT) (high_bits(count)); |
957
|
2745
|
|
|
|
|
|
countlo = (AM_SHORT) (low_bits(count)); |
958
|
|
|
|
|
|
|
|
959
|
|
|
|
|
|
|
/* initialize 0 because it won't be overwritten */ |
960
|
|
|
|
|
|
|
/* |
961
|
|
|
|
|
|
|
* TODO: multiply through p[7] into gangcount[7] |
962
|
|
|
|
|
|
|
* and warn if there's potential overflow |
963
|
|
|
|
|
|
|
*/ |
964
|
2745
|
|
|
|
|
|
gangcount[0] = 0; |
965
|
21960
|
100
|
|
|
|
|
for (i = 0; i < 7; ++i) { |
966
|
19215
|
|
|
|
|
|
gangcount[i] += countlo * p[i]; |
967
|
19215
|
|
|
|
|
|
carry_replace(gangcount, i); |
968
|
|
|
|
|
|
|
} |
969
|
2745
|
|
|
|
|
|
gangcount[7] += countlo * p[7]; |
970
|
|
|
|
|
|
|
|
971
|
|
|
|
|
|
|
/* TODO: why is element 0 not considered here? */ |
972
|
2745
|
50
|
|
|
|
|
if (counthi) { |
973
|
0
|
0
|
|
|
|
|
for (i = 0; i < 6; ++i) { |
974
|
0
|
|
|
|
|
|
gangcount[i + 1] += counthi * p[i]; |
975
|
0
|
|
|
|
|
|
carry(gangcount, i + 1); |
976
|
|
|
|
|
|
|
} |
977
|
|
|
|
|
|
|
} |
978
|
21960
|
100
|
|
|
|
|
for (i = 0; i < 7; ++i) { |
979
|
19215
|
|
|
|
|
|
grandtotal[i] += gangcount[i]; |
980
|
19215
|
|
|
|
|
|
carry(grandtotal, i); |
981
|
|
|
|
|
|
|
} |
982
|
2745
|
|
|
|
|
|
grandtotal[7] += gangcount[7]; |
983
|
|
|
|
|
|
|
|
984
|
2745
|
|
|
|
|
|
tempsv = *hv_fetch(gang, HeKEY(he), NUM_LATTICES * sizeof(AM_SHORT), 1); |
985
|
2745
|
50
|
|
|
|
|
SvUPGRADE(tempsv, SVt_PVNV); |
986
|
2745
|
|
|
|
|
|
sv_setpvn(tempsv, (char *) gangcount, 8 * sizeof(AM_LONG)); |
987
|
2745
|
|
|
|
|
|
normalize(aTHX_ tempsv); |
988
|
2745
|
|
|
|
|
|
normalize(aTHX_ HeVAL(he)); |
989
|
|
|
|
|
|
|
|
990
|
2745
|
|
|
|
|
|
tempsv = *hv_fetch(context_to_class, HeKEY(he), NUM_LATTICES * sizeof(AM_SHORT), 0); |
991
|
2745
|
|
|
|
|
|
this_class = (AM_SHORT) SvUVX(tempsv); |
992
|
2745
|
100
|
|
|
|
|
if (this_class) { |
993
|
2702
|
|
|
|
|
|
AM_LONG *s = (AM_LONG *) SvPVX(sum[this_class]); |
994
|
21616
|
100
|
|
|
|
|
for (i = 0; i < 7; ++i) { |
995
|
18914
|
|
|
|
|
|
*(s + i) += gangcount[i]; |
996
|
18914
|
|
|
|
|
|
carry_pointer(s + i); |
997
|
|
|
|
|
|
|
} |
998
|
|
|
|
|
|
|
} else { |
999
|
43
|
|
|
|
|
|
dataitem = *hv_fetch(itemcontextchainhead, HeKEY(he), NUM_LATTICES * sizeof(AM_SHORT), 0); |
1000
|
2857
|
100
|
|
|
|
|
while (SvIOK(dataitem)) { |
1001
|
112
|
|
|
|
|
|
IV datanum = SvIVX(dataitem); |
1002
|
112
|
|
|
|
|
|
IV ocnum = SvIVX(classes[datanum]); |
1003
|
112
|
|
|
|
|
|
AM_LONG *s = (AM_LONG *) SvPVX(sum[ocnum]); |
1004
|
896
|
100
|
|
|
|
|
for (i = 0; i < 7; ++i) { |
1005
|
784
|
|
|
|
|
|
*(s + i) += p[i]; |
1006
|
784
|
|
|
|
|
|
carry_pointer(s + i); |
1007
|
784
|
|
|
|
|
|
dataitem = itemcontextchain[datanum]; |
1008
|
|
|
|
|
|
|
} |
1009
|
|
|
|
|
|
|
} |
1010
|
|
|
|
|
|
|
} |
1011
|
|
|
|
|
|
|
} |
1012
|
753
|
100
|
|
|
|
|
for (i = 1; i <= num_classes; ++i) normalize(aTHX_ sum[i]) |
1013
|
|
|
|
|
|
|
; |
1014
|
194
|
|
|
|
|
|
tempsv = *hv_fetch(pointers, "grandtotal", 10, 1); |
1015
|
194
|
50
|
|
|
|
|
SvUPGRADE(tempsv, SVt_PVNV); |
1016
|
194
|
|
|
|
|
|
sv_setpvn(tempsv, (char *) grandtotal, 8 * sizeof(AM_LONG)); |
1017
|
194
|
|
|
|
|
|
normalize(aTHX_ tempsv); |
1018
|
|
|
|
|
|
|
|
1019
|
194
|
|
|
|
|
|
Safefree(subcontext); |
1020
|
194
|
|
|
|
|
|
Safefree(subcontext_class); |
1021
|
194
|
|
|
|
|
|
Safefree(intersectlist); |
1022
|
194
|
|
|
|
|
|
Safefree(intersectlist2); |
1023
|
194
|
|
|
|
|
|
Safefree(intersectlist3); |
1024
|
|
|
|
|
|
|
} |