[14] | 1 | /**CFile*********************************************************************** |
---|
| 2 | |
---|
| 3 | FileName [ordPerm.c] |
---|
| 4 | |
---|
| 5 | PackageName [ord] |
---|
| 6 | |
---|
| 7 | Synopsis [Routines to find permutation on latches to minimize MDD size.] |
---|
| 8 | |
---|
| 9 | Author [Serdar Tasiran, Tom Shiple] |
---|
| 10 | |
---|
| 11 | Copyright [Copyright (c) 1994-1996 The Regents of the Univ. of California. |
---|
| 12 | All rights reserved. |
---|
| 13 | |
---|
| 14 | Permission is hereby granted, without written agreement and without license |
---|
| 15 | or royalty fees, to use, copy, modify, and distribute this software and its |
---|
| 16 | documentation for any purpose, provided that the above copyright notice and |
---|
| 17 | the following two paragraphs appear in all copies of this software. |
---|
| 18 | |
---|
| 19 | IN NO EVENT SHALL THE UNIVERSITY OF CALIFORNIA BE LIABLE TO ANY PARTY FOR |
---|
| 20 | DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING OUT |
---|
| 21 | OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN IF THE UNIVERSITY OF |
---|
| 22 | CALIFORNIA HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
---|
| 23 | |
---|
| 24 | THE UNIVERSITY OF CALIFORNIA SPECIFICALLY DISCLAIMS ANY WARRANTIES, |
---|
| 25 | INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND |
---|
| 26 | FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS ON AN |
---|
| 27 | "AS IS" BASIS, AND THE UNIVERSITY OF CALIFORNIA HAS NO OBLIGATION TO PROVIDE |
---|
| 28 | MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.] |
---|
| 29 | |
---|
| 30 | ******************************************************************************/ |
---|
| 31 | |
---|
| 32 | #include "ordInt.h" |
---|
| 33 | |
---|
| 34 | static char rcsid[] UNUSED = "$Id: ordPerm.c,v 1.9 2005/04/16 06:15:25 fabio Exp $"; |
---|
| 35 | |
---|
| 36 | /*---------------------------------------------------------------------------*/ |
---|
| 37 | /* Structure declarations */ |
---|
| 38 | /*---------------------------------------------------------------------------*/ |
---|
| 39 | |
---|
| 40 | typedef struct proc_pair { |
---|
| 41 | long proc_no; |
---|
| 42 | long var_no; |
---|
| 43 | } pair; |
---|
| 44 | |
---|
| 45 | typedef struct proc_struct { |
---|
| 46 | int num_loc_vars; |
---|
| 47 | array_t *o_list; /* of pair */ |
---|
| 48 | array_t *i_list; /* of pair */ |
---|
| 49 | } proc_struct; |
---|
| 50 | |
---|
| 51 | typedef struct Proc_Com_Graph { |
---|
| 52 | int num_vars; |
---|
| 53 | int num_proc; |
---|
| 54 | int *width_list; |
---|
| 55 | proc_struct *proc_list; |
---|
| 56 | int *locations; /* Array of length num_proc used as temporary storage by some |
---|
| 57 | functions */ |
---|
| 58 | int *variables; /* Array of length num_vars used as temporary storage by some |
---|
| 59 | functions */ |
---|
| 60 | |
---|
| 61 | } Proc_Com_Graph; |
---|
| 62 | |
---|
| 63 | |
---|
| 64 | /*---------------------------------------------------------------------------*/ |
---|
| 65 | /* Variable declarations */ |
---|
| 66 | /*---------------------------------------------------------------------------*/ |
---|
| 67 | /**Variable******************************************************************** |
---|
| 68 | |
---|
| 69 | Synopsis [required] |
---|
| 70 | |
---|
| 71 | Description [ |
---|
| 72 | CHOICE = 0 -->Look-ahead of 2 |
---|
| 73 | 1 -->Look-ahead of 1 |
---|
| 74 | 2 -->Touati's heuristic |
---|
| 75 | 3 -->Enter permutation manually |
---|
| 76 | 4 -->Random ordering] |
---|
| 77 | |
---|
| 78 | SeeAlso [optional] |
---|
| 79 | |
---|
| 80 | ******************************************************************************/ |
---|
| 81 | static int number_of_calls; |
---|
| 82 | |
---|
| 83 | |
---|
| 84 | /*---------------------------------------------------------------------------*/ |
---|
| 85 | /* Macro declarations */ |
---|
| 86 | /*---------------------------------------------------------------------------*/ |
---|
| 87 | |
---|
| 88 | |
---|
| 89 | /**AutomaticStart*************************************************************/ |
---|
| 90 | |
---|
| 91 | /*---------------------------------------------------------------------------*/ |
---|
| 92 | /* Static function prototypes */ |
---|
| 93 | /*---------------------------------------------------------------------------*/ |
---|
| 94 | |
---|
| 95 | static lsList NetworkComputeLatchOrder(Ntk_Network_t * network, int verbose); |
---|
| 96 | static int * LatchPermutationCompute(Proc_Com_Graph *PCG, int choice, int verbose); |
---|
| 97 | static double rev_fac(int num_rev_bits); |
---|
| 98 | static double cost_for_cut(Proc_Com_Graph * PCG, int * perm, int end_of_set1); |
---|
| 99 | static void append_best(int * opt_perm, int loc, Proc_Com_Graph * PCG); |
---|
| 100 | static double cost_2(int * opt_perm, int loc, Proc_Com_Graph * PCG); |
---|
| 101 | static long int cost_touati_2(int * opt_perm, int loc, Proc_Com_Graph * PCG); |
---|
| 102 | static void swap(int * opt_perm, int i, int j); |
---|
| 103 | static void append_best_2(int * opt_perm, int loc, Proc_Com_Graph * PCG); |
---|
| 104 | static void append_touati_2(int * opt_perm, int loc, Proc_Com_Graph * PCG); |
---|
| 105 | static void init_heur(Proc_Com_Graph * PCG, int * opt_perm); |
---|
| 106 | static void heur_2(Proc_Com_Graph * PCG, int * opt_perm); |
---|
| 107 | static void heur_touati_la2(Proc_Com_Graph * PCG, int * opt_perm); |
---|
| 108 | static int * opt_proc_order(Proc_Com_Graph * PCG); |
---|
| 109 | static int * opt_touati_order(Proc_Com_Graph * PCG); |
---|
| 110 | static double cost_total(Proc_Com_Graph * PCG, int * perm); |
---|
| 111 | static int * random_permutation(int seed, int n_elts); |
---|
| 112 | |
---|
| 113 | /**AutomaticEnd***************************************************************/ |
---|
| 114 | |
---|
| 115 | |
---|
| 116 | /*---------------------------------------------------------------------------*/ |
---|
| 117 | /* Definition of exported functions */ |
---|
| 118 | /*---------------------------------------------------------------------------*/ |
---|
| 119 | |
---|
| 120 | |
---|
| 121 | /*---------------------------------------------------------------------------*/ |
---|
| 122 | /* Definition of internal functions */ |
---|
| 123 | /*---------------------------------------------------------------------------*/ |
---|
| 124 | |
---|
| 125 | /**Function******************************************************************** |
---|
| 126 | |
---|
| 127 | Synopsis [Computes a total ordering on the combinational outputs of a |
---|
| 128 | network.] |
---|
| 129 | |
---|
| 130 | Description [Computes a total ordering on the combinational outputs of a |
---|
| 131 | network. First, the algorithm orders the data inputs using the permutation |
---|
| 132 | algorithm (Aziz et al, "BDD Variable Ordering for Interacting FSMs", DAC |
---|
| 133 | 1994, p. 283). Then, it orders the remaining combinational outputs in order |
---|
| 134 | of decreasing depth. Finally, the second list is appended to the first.] |
---|
| 135 | |
---|
| 136 | SideEffects [] |
---|
| 137 | |
---|
| 138 | SeeAlso [OrdNetworkComputeNodeDepths] |
---|
| 139 | |
---|
| 140 | ******************************************************************************/ |
---|
| 141 | lsList |
---|
| 142 | OrdNetworkOrderRootsByPerm( |
---|
| 143 | Ntk_Network_t *network, |
---|
| 144 | int verbose) |
---|
| 145 | { |
---|
| 146 | lsGen gen; |
---|
| 147 | Ntk_Node_t *node; |
---|
| 148 | lsList dataInputs; |
---|
| 149 | st_table *processedTable = st_init_table(st_ptrcmp, st_ptrhash); |
---|
| 150 | lsList otherCombOuts = lsCreate(); |
---|
| 151 | |
---|
| 152 | /* |
---|
| 153 | * Create a list of all combinational outputs that are not data inputs to |
---|
| 154 | * latches. A node can drive more than one latch data input, latch initial |
---|
| 155 | * input, or primary output. Use a hash table to ensure that no node appears |
---|
| 156 | * twice in the list. |
---|
| 157 | */ |
---|
| 158 | Ntk_NetworkForEachCombOutput(network, gen, node) { |
---|
| 159 | if (!Ntk_NodeTestIsLatchDataInput(node)) { |
---|
| 160 | OrdNodeAddToList(otherCombOuts, processedTable, node); |
---|
| 161 | } |
---|
| 162 | } |
---|
| 163 | st_free_table(processedTable); |
---|
| 164 | |
---|
| 165 | /* Compute depth of all roots in otherCombOuts. */ |
---|
| 166 | OrdNetworkComputeNodeDepths(network, otherCombOuts); |
---|
| 167 | |
---|
| 168 | /* Sort otherCombOuts based on depth. */ |
---|
| 169 | lsSort(otherCombOuts, OrdNodesFromListCompareDepth); |
---|
| 170 | |
---|
| 171 | /* Compute order on dataInputs roots using the permutation method. */ |
---|
| 172 | dataInputs = NetworkComputeLatchOrder(network, verbose); |
---|
| 173 | |
---|
| 174 | /* Add the otherCombOuts list to the end of the dataInputs list. */ |
---|
| 175 | Ord_ListAppendList(dataInputs, otherCombOuts); |
---|
| 176 | (void) lsDestroy(otherCombOuts, (void (*) (lsGeneric)) NULL); |
---|
| 177 | |
---|
| 178 | return dataInputs; |
---|
| 179 | } |
---|
| 180 | |
---|
| 181 | |
---|
| 182 | /*---------------------------------------------------------------------------*/ |
---|
| 183 | /* Definition of static functions */ |
---|
| 184 | /*---------------------------------------------------------------------------*/ |
---|
| 185 | |
---|
| 186 | /**Function******************************************************************** |
---|
| 187 | |
---|
| 188 | Synopsis [Computes an ordering on the latches.] |
---|
| 189 | |
---|
| 190 | Description [Computes an ordering on the latches, to be used for variable |
---|
| 191 | ordering. Returns a list of the corresponding latch data inputs (Ntk_Node_t |
---|
| 192 | *) giving the ordering. Implements the algorithm given in Aziz et al, "BDD |
---|
| 193 | Variable Ordering for Interacting FSMs", DAC 1994, p. 283.] |
---|
| 194 | |
---|
| 195 | SideEffects [Creates a file in /tmp.] |
---|
| 196 | |
---|
| 197 | SeeAlso [NetworkOrderNodes] |
---|
| 198 | |
---|
| 199 | ******************************************************************************/ |
---|
| 200 | static lsList |
---|
| 201 | NetworkComputeLatchOrder( |
---|
| 202 | Ntk_Network_t * network, |
---|
| 203 | int verbose) |
---|
| 204 | { |
---|
| 205 | int i, j; |
---|
| 206 | lsGen listGen; |
---|
| 207 | st_generator *stGen; |
---|
| 208 | Ntk_Node_t *latch; |
---|
| 209 | lsList tfoLatchList; |
---|
| 210 | int *latchOrderArray; |
---|
| 211 | long count = 0L; |
---|
| 212 | st_table *idToLatch = st_init_table(st_numcmp, st_numhash); |
---|
| 213 | int numLatches = Ntk_NetworkReadNumLatches(network); |
---|
| 214 | st_table *latchDependencies = Ntk_NetworkComputeLatchDependencies(network); |
---|
| 215 | lsList latchOrderList = lsCreate(); |
---|
| 216 | int latchOrderingMode = 1; /* FIX: make an option */ |
---|
| 217 | double convConstant = log10(2.0); |
---|
| 218 | Proc_Com_Graph *PCG; |
---|
| 219 | pair *cur_pair; |
---|
| 220 | |
---|
| 221 | /* |
---|
| 222 | * Assign unique integer between 0 and numLatches-1 to each latch. Store |
---|
| 223 | * this correspondance in the idToLatch hash table. |
---|
| 224 | * (NOTE: may be sensitive to ordering in memory.) |
---|
| 225 | */ |
---|
| 226 | Ntk_NetworkForEachLatch(network, listGen, latch) { |
---|
| 227 | Ntk_NodeSetUndef(latch, (void *) count); |
---|
| 228 | st_insert(idToLatch, (char *) count, (char *) latch); |
---|
| 229 | count++; |
---|
| 230 | } |
---|
| 231 | |
---|
| 232 | /* Create a Proc_Com_Graph and write the communications |
---|
| 233 | * structure of the network into it |
---|
| 234 | */ |
---|
| 235 | |
---|
| 236 | PCG = ALLOC(Proc_Com_Graph, 1); |
---|
| 237 | PCG->num_proc = numLatches; /* SER (void) fprintf(pcgFile, "%d\n", numLatches); */ |
---|
| 238 | PCG->num_vars = numLatches; /* SER (void) fprintf(pcgFile, "%d\n", numLatches); */ |
---|
| 239 | |
---|
| 240 | PCG->proc_list = ALLOC(proc_struct, PCG->num_proc); |
---|
| 241 | PCG->width_list= ALLOC(int, PCG->num_vars); |
---|
| 242 | PCG->locations = ALLOC(int, PCG->num_proc); |
---|
| 243 | PCG->variables = ALLOC(int, PCG->num_vars); |
---|
| 244 | |
---|
| 245 | for(i=0;i< PCG->num_proc ; i++) { |
---|
| 246 | PCG->proc_list[i].o_list= array_alloc(pair *, 0); |
---|
| 247 | PCG->proc_list[i].i_list= array_alloc(pair *, 0); |
---|
| 248 | }; |
---|
| 249 | |
---|
| 250 | /* |
---|
| 251 | * For each latch/tfoLatchList pair, write information about the latch, and |
---|
| 252 | * write the latches to which the latch fanouts. Finish the entry with "%%". |
---|
| 253 | * Here is the general format for a process: |
---|
| 254 | * process # list of local vars to process # |
---|
| 255 | * fanout-latch : local-var-fanning-to width-of-local-var / |
---|
| 256 | * .... |
---|
| 257 | * fanout-latch : local-var-fanning-to width-of-local-var / %% |
---|
| 258 | * |
---|
| 259 | * When this format is specialized to case where each process is a single |
---|
| 260 | * latch, then the format looks like: |
---|
| 261 | * latch # latch # |
---|
| 262 | * fanout-latch : latch width-of-latch / |
---|
| 263 | * .... |
---|
| 264 | * fanout-latch : latch width-of-latch / %% |
---|
| 265 | */ |
---|
| 266 | st_foreach_item(latchDependencies, stGen, &latch, &tfoLatchList) { |
---|
| 267 | Ntk_Node_t *tfoLatch; |
---|
| 268 | int varWidth; |
---|
| 269 | long var_no, fr_no, to_no; |
---|
| 270 | |
---|
| 271 | |
---|
| 272 | /* |
---|
| 273 | * Write the latch id and the cardinality of the latch variable |
---|
| 274 | * domain. |
---|
| 275 | */ |
---|
| 276 | fr_no = (long) Ntk_NodeReadUndef(latch); |
---|
| 277 | PCG->proc_list[fr_no].num_loc_vars = 1; |
---|
| 278 | |
---|
| 279 | |
---|
| 280 | /* |
---|
| 281 | * For each transitive fanout latch, write the tfoLatch id, the latch |
---|
| 282 | * id, and the width of the latch variable (assumes minimum |
---|
| 283 | * encoding). Width is ceiling(log2(number of values in domain)). |
---|
| 284 | */ |
---|
| 285 | varWidth = (int) ceil(log10 |
---|
| 286 | ((double)Var_VariableReadNumValues(Ntk_NodeReadVariable(latch))) |
---|
| 287 | / convConstant); |
---|
| 288 | lsForEachItem(tfoLatchList, listGen, tfoLatch) { |
---|
| 289 | to_no = (long) Ntk_NodeReadUndef(tfoLatch); |
---|
| 290 | var_no = (long) Ntk_NodeReadUndef(latch); |
---|
| 291 | |
---|
| 292 | cur_pair = ALLOC(pair,1); |
---|
| 293 | cur_pair->proc_no = to_no; |
---|
| 294 | cur_pair->var_no = var_no; |
---|
| 295 | array_insert_last(pair *, PCG->proc_list[fr_no].o_list, cur_pair); |
---|
| 296 | |
---|
| 297 | cur_pair = ALLOC(pair,1); |
---|
| 298 | cur_pair->proc_no = fr_no; |
---|
| 299 | cur_pair->var_no = var_no; |
---|
| 300 | array_insert_last(pair *, PCG->proc_list[to_no].i_list, cur_pair); |
---|
| 301 | |
---|
| 302 | PCG->width_list[var_no] = varWidth; |
---|
| 303 | } |
---|
| 304 | } |
---|
| 305 | |
---|
| 306 | /* |
---|
| 307 | * We don't need the latchDependencies table anymore. Free the list stored |
---|
| 308 | * at each value, and then free the table itself. |
---|
| 309 | */ |
---|
| 310 | st_foreach_item(latchDependencies, stGen, &latch, &tfoLatchList) { |
---|
| 311 | (void) lsDestroy(tfoLatchList, (void (*) (lsGeneric)) NULL); |
---|
| 312 | } |
---|
| 313 | st_free_table(latchDependencies); |
---|
| 314 | |
---|
| 315 | /* |
---|
| 316 | * Compute the order on the latches. This is returned as an array of |
---|
| 317 | * integers, where the latch whose id is latchOrderArray[i] is ordered in |
---|
| 318 | * the ith position of latchOrderList. Note that the returned list actually |
---|
| 319 | * contains the latch data inputs, not the latches themselves. |
---|
| 320 | */ |
---|
| 321 | latchOrderArray = LatchPermutationCompute(PCG, latchOrderingMode, verbose); |
---|
| 322 | |
---|
| 323 | for (i=0; i < PCG->num_proc; i++) { |
---|
| 324 | for (j=0; j<array_n(PCG->proc_list[i].o_list); j++) { |
---|
| 325 | cur_pair = array_fetch(pair *, PCG->proc_list[i].o_list, j); |
---|
| 326 | FREE(cur_pair); |
---|
| 327 | } |
---|
| 328 | |
---|
| 329 | for (j=0; j<array_n(PCG->proc_list[i].i_list); j++) { |
---|
| 330 | cur_pair = array_fetch(pair *, PCG->proc_list[i].i_list, j); |
---|
| 331 | FREE(cur_pair); |
---|
| 332 | } |
---|
| 333 | |
---|
| 334 | array_free(PCG->proc_list[i].o_list); |
---|
| 335 | array_free(PCG->proc_list[i].i_list); |
---|
| 336 | } |
---|
| 337 | |
---|
| 338 | FREE(PCG->proc_list); |
---|
| 339 | FREE(PCG->width_list); |
---|
| 340 | FREE(PCG->variables); |
---|
| 341 | FREE(PCG->locations); |
---|
| 342 | FREE(PCG); |
---|
| 343 | |
---|
| 344 | for (i = 0; i < numLatches; i++) { |
---|
| 345 | int status UNUSED = st_lookup(idToLatch, |
---|
| 346 | (char *) (long) (latchOrderArray[i]), |
---|
| 347 | &latch); |
---|
| 348 | assert(status); |
---|
| 349 | (void) lsNewEnd(latchOrderList, (lsGeneric) Ntk_LatchReadDataInput(latch), LS_NH); |
---|
| 350 | } |
---|
| 351 | |
---|
| 352 | /* |
---|
| 353 | * We are done with idToLatch and latchOrderArray. |
---|
| 354 | */ |
---|
| 355 | st_free_table(idToLatch); |
---|
| 356 | FREE(latchOrderArray); |
---|
| 357 | |
---|
| 358 | return (latchOrderList); |
---|
| 359 | } |
---|
| 360 | |
---|
| 361 | |
---|
| 362 | /**Function******************************************************************** |
---|
| 363 | |
---|
| 364 | Synopsis [required] |
---|
| 365 | |
---|
| 366 | Description [optional] |
---|
| 367 | |
---|
| 368 | SideEffects [required] |
---|
| 369 | |
---|
| 370 | SeeAlso [optional] |
---|
| 371 | |
---|
| 372 | ******************************************************************************/ |
---|
| 373 | static int * |
---|
| 374 | LatchPermutationCompute( |
---|
| 375 | Proc_Com_Graph *PCG, |
---|
| 376 | int choice, |
---|
| 377 | int verbose) |
---|
| 378 | { |
---|
| 379 | int *opt_perm; |
---|
| 380 | |
---|
| 381 | if (choice == 0) { |
---|
| 382 | opt_perm = opt_proc_order(PCG); |
---|
| 383 | } |
---|
| 384 | else if (choice == 1) { |
---|
| 385 | opt_perm = ALLOC(int, PCG->num_proc); |
---|
| 386 | init_heur(PCG, opt_perm); |
---|
| 387 | } |
---|
| 388 | else if (choice == 2) { |
---|
| 389 | opt_perm = opt_touati_order(PCG); |
---|
| 390 | } |
---|
| 391 | else if (choice == 3) { |
---|
| 392 | opt_perm = random_permutation(7,PCG->num_vars); |
---|
| 393 | } |
---|
| 394 | else { |
---|
| 395 | fail("Unknown Option for Computing Latch Permutation"); |
---|
| 396 | } |
---|
| 397 | |
---|
| 398 | if (verbose > 1) { |
---|
| 399 | (void) fprintf(vis_stdout, "TOTAL COST= %e \n",cost_total(PCG,opt_perm)); |
---|
| 400 | } |
---|
| 401 | |
---|
| 402 | return opt_perm; |
---|
| 403 | } |
---|
| 404 | |
---|
| 405 | |
---|
| 406 | /**Function******************************************************************** |
---|
| 407 | |
---|
| 408 | Synopsis [required] |
---|
| 409 | |
---|
| 410 | Description [optional] |
---|
| 411 | |
---|
| 412 | SideEffects [required] |
---|
| 413 | |
---|
| 414 | SeeAlso [optional] |
---|
| 415 | |
---|
| 416 | ******************************************************************************/ |
---|
| 417 | static double |
---|
| 418 | rev_fac( |
---|
| 419 | int num_rev_bits) |
---|
| 420 | { |
---|
| 421 | double result; |
---|
| 422 | |
---|
| 423 | result = pow( ((double) 2.0),((double) num_rev_bits) ); |
---|
| 424 | |
---|
| 425 | return result; |
---|
| 426 | } |
---|
| 427 | |
---|
| 428 | |
---|
| 429 | /**Function******************************************************************** |
---|
| 430 | |
---|
| 431 | Synopsis [required] |
---|
| 432 | |
---|
| 433 | Description [Calculates the cost for a single cut that divides the |
---|
| 434 | permutation into 2.] |
---|
| 435 | |
---|
| 436 | SideEffects [required] |
---|
| 437 | |
---|
| 438 | SeeAlso [optional] |
---|
| 439 | |
---|
| 440 | ******************************************************************************/ |
---|
| 441 | static double |
---|
| 442 | cost_for_cut( |
---|
| 443 | Proc_Com_Graph * PCG, |
---|
| 444 | int * perm, |
---|
| 445 | int end_of_set1) |
---|
| 446 | { |
---|
| 447 | int i, j; |
---|
| 448 | double result; |
---|
| 449 | int forw = 0, rev = 0; |
---|
| 450 | pair *cur_pair; |
---|
| 451 | |
---|
| 452 | for (i=0; i<=end_of_set1; i++) { |
---|
| 453 | PCG->locations[perm[i]]=1; |
---|
| 454 | } |
---|
| 455 | |
---|
| 456 | for (i=end_of_set1+1; i<PCG->num_proc; i++) { |
---|
| 457 | PCG->locations[perm[i]]=2; |
---|
| 458 | } |
---|
| 459 | |
---|
| 460 | for (i=0; i<PCG->num_vars; i++) { |
---|
| 461 | PCG->variables[i]=0; |
---|
| 462 | } |
---|
| 463 | |
---|
| 464 | /* Count the # of forward crossing bits */ |
---|
| 465 | /* for (i=0; i<=end_of_set1; i++) { |
---|
| 466 | st_foreach_item(PCG->proc_list[perm[i]].o_list, |
---|
| 467 | gen,(char **)&bit_no,(char **)&to_proc) { |
---|
| 468 | if (PCG->locations[to_proc] ==2) { |
---|
| 469 | PCG->variables[bit_no]++ ; |
---|
| 470 | } |
---|
| 471 | } |
---|
| 472 | } |
---|
| 473 | */ |
---|
| 474 | for (i=0; i<=end_of_set1; i++) |
---|
| 475 | for (j=0; j<array_n(PCG->proc_list[perm[i]].o_list); j++) { |
---|
| 476 | cur_pair = array_fetch(pair *, PCG->proc_list[perm[i]].o_list, j); |
---|
| 477 | if (PCG->locations[cur_pair->proc_no] ==2) { |
---|
| 478 | PCG->variables[cur_pair->var_no]++ ; |
---|
| 479 | } |
---|
| 480 | } |
---|
| 481 | |
---|
| 482 | |
---|
| 483 | for (i=0; i<PCG->num_vars; i++) { |
---|
| 484 | if (PCG->variables[i] > 0) { |
---|
| 485 | ++forw ; |
---|
| 486 | } |
---|
| 487 | } |
---|
| 488 | |
---|
| 489 | for (i=0; i<PCG->num_vars; i++) { |
---|
| 490 | PCG->variables[i]=0; |
---|
| 491 | } |
---|
| 492 | |
---|
| 493 | /* Count the # of reverse crossing bits */ |
---|
| 494 | /* for (i=end_of_set1+1; i<PCG->num_proc; i++) { |
---|
| 495 | st_foreach_item(PCG->proc_list[perm[i]].o_list, |
---|
| 496 | gen,(char **)&bit_no,(char **)&to_proc) { |
---|
| 497 | if (PCG->locations[to_proc] ==1) { |
---|
| 498 | PCG->variables[bit_no]++; |
---|
| 499 | } |
---|
| 500 | } |
---|
| 501 | } |
---|
| 502 | */ |
---|
| 503 | for (i=end_of_set1+1; i<PCG->num_proc; i++) |
---|
| 504 | for (j=0; j<array_n(PCG->proc_list[perm[i]].o_list); j++) { |
---|
| 505 | cur_pair = array_fetch(pair *, PCG->proc_list[perm[i]].o_list, j); |
---|
| 506 | if (PCG->locations[cur_pair->proc_no] ==1) { |
---|
| 507 | PCG->variables[cur_pair->var_no]++ ; |
---|
| 508 | } |
---|
| 509 | } |
---|
| 510 | |
---|
| 511 | |
---|
| 512 | for (i=0; i<PCG->num_vars; i++) { |
---|
| 513 | if (PCG->variables[i] > 0) { |
---|
| 514 | ++rev; |
---|
| 515 | } |
---|
| 516 | } |
---|
| 517 | |
---|
| 518 | { |
---|
| 519 | double rev_cost = rev_fac(rev); |
---|
| 520 | result = ((double) forw) + rev_cost; |
---|
| 521 | number_of_calls++; |
---|
| 522 | } |
---|
| 523 | return result; |
---|
| 524 | } |
---|
| 525 | |
---|
| 526 | |
---|
| 527 | /**Function******************************************************************** |
---|
| 528 | |
---|
| 529 | Synopsis [required] |
---|
| 530 | |
---|
| 531 | Description [Adds the next best element in the greedy sense as described for |
---|
| 532 | init_heur below.] |
---|
| 533 | |
---|
| 534 | SideEffects [required] |
---|
| 535 | |
---|
| 536 | SeeAlso [optional] |
---|
| 537 | |
---|
| 538 | ******************************************************************************/ |
---|
| 539 | static void |
---|
| 540 | append_best( |
---|
| 541 | int * opt_perm, |
---|
| 542 | int loc, |
---|
| 543 | Proc_Com_Graph * PCG) |
---|
| 544 | { |
---|
| 545 | int i, temp, bestswap, cur_fanout, best_fanout; |
---|
| 546 | double cur_cost, best_cost; |
---|
| 547 | int j; |
---|
| 548 | pair *cur_pair; |
---|
| 549 | |
---|
| 550 | best_cost = cost_for_cut(PCG, opt_perm, loc); |
---|
| 551 | cur_cost = best_cost; |
---|
| 552 | bestswap = loc; |
---|
| 553 | |
---|
| 554 | best_fanout = 100000; |
---|
| 555 | for(i=loc+1; i<PCG->num_proc; i++) { |
---|
| 556 | temp = opt_perm[loc]; |
---|
| 557 | opt_perm[loc] = opt_perm[i]; |
---|
| 558 | opt_perm[i] = temp; |
---|
| 559 | |
---|
| 560 | cur_cost = cost_for_cut(PCG, opt_perm, loc); |
---|
| 561 | |
---|
| 562 | if ( cur_cost == best_cost ) { |
---|
| 563 | cur_fanout = 0; |
---|
| 564 | |
---|
| 565 | for ( j = 0 ; j <= loc ; j++ ) { |
---|
| 566 | PCG->locations[opt_perm[j]] = 1; |
---|
| 567 | } |
---|
| 568 | |
---|
| 569 | for ( j = loc + 1 ; j < PCG->num_proc ; j++ ) { |
---|
| 570 | PCG->locations[opt_perm[j]] = 2; |
---|
| 571 | } |
---|
| 572 | |
---|
| 573 | /* st_foreach_item( PCG->proc_list[opt_perm[i]].o_list, |
---|
| 574 | stgen, ( char ** ) &bit_no, ( char ** ) &to_proc ) { |
---|
| 575 | if ( PCG->locations[to_proc] == 2 ) { |
---|
| 576 | cur_fanout++; |
---|
| 577 | } |
---|
| 578 | } |
---|
| 579 | */ |
---|
| 580 | for (j=0; j<array_n(PCG->proc_list[opt_perm[i]].o_list); j++) { |
---|
| 581 | cur_pair = array_fetch(pair *, PCG->proc_list[opt_perm[i]].o_list, j); |
---|
| 582 | if (PCG->locations[cur_pair->proc_no] ==2) |
---|
| 583 | cur_fanout++; |
---|
| 584 | } |
---|
| 585 | |
---|
| 586 | |
---|
| 587 | |
---|
| 588 | |
---|
| 589 | if ( cur_fanout < best_fanout ) { |
---|
| 590 | best_fanout = cur_fanout; |
---|
| 591 | bestswap = i; |
---|
| 592 | } |
---|
| 593 | } |
---|
| 594 | |
---|
| 595 | if (cur_cost < best_cost) { |
---|
| 596 | best_cost = cur_cost; |
---|
| 597 | bestswap = i; |
---|
| 598 | } |
---|
| 599 | |
---|
| 600 | temp = opt_perm[loc]; |
---|
| 601 | opt_perm[loc] = opt_perm[i]; |
---|
| 602 | opt_perm[i] = temp; |
---|
| 603 | } |
---|
| 604 | |
---|
| 605 | temp = opt_perm[loc]; |
---|
| 606 | opt_perm[loc] = opt_perm[bestswap]; |
---|
| 607 | opt_perm[bestswap] = temp; |
---|
| 608 | } |
---|
| 609 | |
---|
| 610 | |
---|
| 611 | /**Function******************************************************************** |
---|
| 612 | |
---|
| 613 | Synopsis [required] |
---|
| 614 | |
---|
| 615 | Description [optional] |
---|
| 616 | |
---|
| 617 | SideEffects [required] |
---|
| 618 | |
---|
| 619 | SeeAlso [optional] |
---|
| 620 | |
---|
| 621 | ******************************************************************************/ |
---|
| 622 | static double |
---|
| 623 | cost_2( |
---|
| 624 | int * opt_perm, |
---|
| 625 | int loc, |
---|
| 626 | Proc_Com_Graph * PCG) |
---|
| 627 | { |
---|
| 628 | |
---|
| 629 | double cost_0 = cost_for_cut(PCG, opt_perm, loc); |
---|
| 630 | double cost_1 = cost_for_cut(PCG, opt_perm, loc+1); |
---|
| 631 | double result = cost_0 + cost_1; |
---|
| 632 | |
---|
| 633 | return result; |
---|
| 634 | |
---|
| 635 | } |
---|
| 636 | |
---|
| 637 | |
---|
| 638 | /**Function******************************************************************** |
---|
| 639 | |
---|
| 640 | Synopsis [required] |
---|
| 641 | |
---|
| 642 | Description [optional] |
---|
| 643 | |
---|
| 644 | SideEffects [required] |
---|
| 645 | |
---|
| 646 | SeeAlso [optional] |
---|
| 647 | |
---|
| 648 | ******************************************************************************/ |
---|
| 649 | static long int |
---|
| 650 | cost_touati_2( |
---|
| 651 | int * opt_perm, |
---|
| 652 | int loc, |
---|
| 653 | Proc_Com_Graph * PCG) |
---|
| 654 | { |
---|
| 655 | int i, j; |
---|
| 656 | long int result; |
---|
| 657 | pair *cur_pair; |
---|
| 658 | |
---|
| 659 | |
---|
| 660 | for (i=0; i<loc; i++) { |
---|
| 661 | PCG->locations[opt_perm[i]]=1; |
---|
| 662 | } |
---|
| 663 | |
---|
| 664 | for (i=loc+1; i<PCG->num_proc; i++) { |
---|
| 665 | PCG->locations[opt_perm[i]]=0; |
---|
| 666 | } |
---|
| 667 | |
---|
| 668 | for (i=0; i<PCG->num_vars; i++) |
---|
| 669 | PCG->variables[i]=0; |
---|
| 670 | |
---|
| 671 | /* st_foreach_item(PCG->proc_list[opt_perm[loc]].i_list, |
---|
| 672 | gen,(char **)&bit_no,(char **)&from_proc) { |
---|
| 673 | if (PCG->locations[from_proc] == 0) { |
---|
| 674 | PCG->variables[bit_no]++ ; |
---|
| 675 | } |
---|
| 676 | } |
---|
| 677 | */ |
---|
| 678 | |
---|
| 679 | for (j=0; j<array_n(PCG->proc_list[opt_perm[loc]].i_list); j++) { |
---|
| 680 | cur_pair = array_fetch(pair *, PCG->proc_list[opt_perm[loc]].i_list, j); |
---|
| 681 | if (PCG->locations[cur_pair->proc_no] ==0) { |
---|
| 682 | PCG->variables[cur_pair->var_no]++ ; |
---|
| 683 | } |
---|
| 684 | } |
---|
| 685 | |
---|
| 686 | result = 0; |
---|
| 687 | for (i=0; i<PCG->num_vars; i++) { |
---|
| 688 | if (PCG->variables[i] != 0) { |
---|
| 689 | result++; |
---|
| 690 | } |
---|
| 691 | } |
---|
| 692 | |
---|
| 693 | for (i=0; i<PCG->num_vars; i++) |
---|
| 694 | PCG->variables[i]=0; |
---|
| 695 | |
---|
| 696 | /* st_foreach_item(PCG->proc_list[opt_perm[loc+1]].i_list, |
---|
| 697 | gen,(char **)&bit_no,(char **)&from_proc) { |
---|
| 698 | if (PCG->locations[from_proc] == 0) { |
---|
| 699 | PCG->variables[bit_no]++ ; |
---|
| 700 | } |
---|
| 701 | } |
---|
| 702 | */ |
---|
| 703 | |
---|
| 704 | for (j=0; j<array_n(PCG->proc_list[opt_perm[loc+1]].i_list); j++) { |
---|
| 705 | cur_pair = array_fetch(pair *, PCG->proc_list[opt_perm[loc+1]].i_list, j); |
---|
| 706 | if (PCG->locations[cur_pair->proc_no] ==0) { |
---|
| 707 | PCG->variables[cur_pair->var_no]++ ; |
---|
| 708 | } |
---|
| 709 | } |
---|
| 710 | |
---|
| 711 | for (i=0; i<PCG->num_vars; i++) { |
---|
| 712 | if (PCG->variables[i] != 0) { |
---|
| 713 | result++; |
---|
| 714 | } |
---|
| 715 | } |
---|
| 716 | |
---|
| 717 | return result; |
---|
| 718 | } |
---|
| 719 | |
---|
| 720 | |
---|
| 721 | /**Function******************************************************************** |
---|
| 722 | |
---|
| 723 | Synopsis [required] |
---|
| 724 | |
---|
| 725 | Description [optional] |
---|
| 726 | |
---|
| 727 | SideEffects [required] |
---|
| 728 | |
---|
| 729 | SeeAlso [optional] |
---|
| 730 | |
---|
| 731 | ******************************************************************************/ |
---|
| 732 | static void |
---|
| 733 | swap( |
---|
| 734 | int * opt_perm, |
---|
| 735 | int i, |
---|
| 736 | int j) |
---|
| 737 | { |
---|
| 738 | int temp; |
---|
| 739 | |
---|
| 740 | temp = opt_perm[i]; |
---|
| 741 | opt_perm[i] = opt_perm[j]; |
---|
| 742 | opt_perm[j] = temp; |
---|
| 743 | } |
---|
| 744 | |
---|
| 745 | |
---|
| 746 | /**Function******************************************************************** |
---|
| 747 | |
---|
| 748 | Synopsis [required] |
---|
| 749 | |
---|
| 750 | Description [optional] |
---|
| 751 | |
---|
| 752 | SideEffects [required] |
---|
| 753 | |
---|
| 754 | SeeAlso [optional] |
---|
| 755 | |
---|
| 756 | ******************************************************************************/ |
---|
| 757 | static void |
---|
| 758 | append_best_2( |
---|
| 759 | int * opt_perm, |
---|
| 760 | int loc, |
---|
| 761 | Proc_Com_Graph * PCG) |
---|
| 762 | { |
---|
| 763 | double best_cost, cur_cost; |
---|
| 764 | int i,j; |
---|
| 765 | int best[2]; |
---|
| 766 | |
---|
| 767 | |
---|
| 768 | best_cost = cost_2(opt_perm, loc, PCG); |
---|
| 769 | cur_cost = best_cost; |
---|
| 770 | best[0] = loc; |
---|
| 771 | best[1] = loc+1; |
---|
| 772 | |
---|
| 773 | for(i=loc; i<PCG->num_proc; i++) { |
---|
| 774 | swap(opt_perm,loc,i); |
---|
| 775 | for(j=loc+1; j<PCG->num_proc; j++) { |
---|
| 776 | swap(opt_perm,loc+1,j); |
---|
| 777 | cur_cost = cost_2(opt_perm, loc, PCG); |
---|
| 778 | if (cur_cost < best_cost) { |
---|
| 779 | best_cost = cur_cost; |
---|
| 780 | best[0] = i; |
---|
| 781 | best[1] = j; |
---|
| 782 | }; |
---|
| 783 | swap(opt_perm,loc+1,j); |
---|
| 784 | } |
---|
| 785 | swap(opt_perm,loc,i); |
---|
| 786 | }; |
---|
| 787 | |
---|
| 788 | swap(opt_perm,loc,best[0]); |
---|
| 789 | swap(opt_perm,loc+1,best[1]); |
---|
| 790 | } |
---|
| 791 | |
---|
| 792 | |
---|
| 793 | /**Function******************************************************************** |
---|
| 794 | |
---|
| 795 | Synopsis [required] |
---|
| 796 | |
---|
| 797 | Description [optional] |
---|
| 798 | |
---|
| 799 | SideEffects [required] |
---|
| 800 | |
---|
| 801 | SeeAlso [optional] |
---|
| 802 | |
---|
| 803 | ******************************************************************************/ |
---|
| 804 | static void |
---|
| 805 | append_touati_2( |
---|
| 806 | int * opt_perm, |
---|
| 807 | int loc, |
---|
| 808 | Proc_Com_Graph * PCG) |
---|
| 809 | { |
---|
| 810 | long int best_cost, cur_cost; |
---|
| 811 | int i,j; |
---|
| 812 | int best[2]; |
---|
| 813 | |
---|
| 814 | /* |
---|
| 815 | * FIX: this is the same function as append_best_2, except that |
---|
| 816 | * cost_touati_2 is used rather than cost_2; just pass the cost function in |
---|
| 817 | * as a parameter. |
---|
| 818 | */ |
---|
| 819 | |
---|
| 820 | best_cost = cost_touati_2(opt_perm, loc, PCG); |
---|
| 821 | cur_cost = best_cost; |
---|
| 822 | best[0] = loc; |
---|
| 823 | best[1] = loc+1; |
---|
| 824 | |
---|
| 825 | for(i=loc; i<PCG->num_proc; i++) { |
---|
| 826 | swap(opt_perm,loc,i); |
---|
| 827 | for(j=loc+1; j<PCG->num_proc; j++) { |
---|
| 828 | swap(opt_perm,loc+1,j); |
---|
| 829 | cur_cost = cost_touati_2(opt_perm, loc, PCG); |
---|
| 830 | if (cur_cost < best_cost) { |
---|
| 831 | best_cost = cur_cost; |
---|
| 832 | best[0] = i; |
---|
| 833 | best[1] = j; |
---|
| 834 | }; |
---|
| 835 | swap(opt_perm,loc+1,j); |
---|
| 836 | } |
---|
| 837 | swap(opt_perm,loc,i); |
---|
| 838 | }; |
---|
| 839 | |
---|
| 840 | swap(opt_perm,loc,best[0]); |
---|
| 841 | swap(opt_perm,loc+1,best[1]); |
---|
| 842 | } |
---|
| 843 | |
---|
| 844 | |
---|
| 845 | /**Function******************************************************************** |
---|
| 846 | |
---|
| 847 | Synopsis [required] |
---|
| 848 | |
---|
| 849 | Description [Returns an initial guess for the optimum permutation by forming |
---|
| 850 | it one element at a time at each step choosing the element that minimizes |
---|
| 851 | cost_for_cut for the set obtained by adding it to the first part of the |
---|
| 852 | permutation, writes the permutation to opt_perm.] |
---|
| 853 | |
---|
| 854 | SideEffects [required] |
---|
| 855 | |
---|
| 856 | SeeAlso [optional] |
---|
| 857 | |
---|
| 858 | ******************************************************************************/ |
---|
| 859 | static void |
---|
| 860 | init_heur( |
---|
| 861 | Proc_Com_Graph * PCG, |
---|
| 862 | int * opt_perm) |
---|
| 863 | { |
---|
| 864 | int i; |
---|
| 865 | |
---|
| 866 | for(i=0; i<PCG->num_proc; i++) { |
---|
| 867 | opt_perm[i] = i; |
---|
| 868 | } |
---|
| 869 | |
---|
| 870 | |
---|
| 871 | |
---|
| 872 | /*printf("\nEntering ordering code:\n");*/ |
---|
| 873 | for(i=0; i<PCG->num_proc; i++) { |
---|
| 874 | /*printf("Iteration\t:\t%d(%d)\n", i, PCG->num_proc );*/ |
---|
| 875 | append_best(opt_perm, i, PCG); |
---|
| 876 | } |
---|
| 877 | /*printf("\n\t--- Total number of calls = %d ---\n", number_of_calls );*/ |
---|
| 878 | } |
---|
| 879 | |
---|
| 880 | |
---|
| 881 | /**Function******************************************************************** |
---|
| 882 | |
---|
| 883 | Synopsis [required] |
---|
| 884 | |
---|
| 885 | Description [optional] |
---|
| 886 | |
---|
| 887 | SideEffects [required] |
---|
| 888 | |
---|
| 889 | SeeAlso [optional] |
---|
| 890 | |
---|
| 891 | ******************************************************************************/ |
---|
| 892 | static void |
---|
| 893 | heur_2( |
---|
| 894 | Proc_Com_Graph * PCG, |
---|
| 895 | int * opt_perm) |
---|
| 896 | { |
---|
| 897 | |
---|
| 898 | int i; |
---|
| 899 | |
---|
| 900 | for(i=0; i<PCG->num_proc; i++) { |
---|
| 901 | opt_perm[i] = i; |
---|
| 902 | } |
---|
| 903 | |
---|
| 904 | for(i=0; i<(2*floor(((double)PCG->num_proc) / 2.0 )); i = i+2) { |
---|
| 905 | /*printf("Iteration\t:\t%d(%d)\n", i, PCG->num_proc );*/ |
---|
| 906 | append_best_2(opt_perm,i,PCG); |
---|
| 907 | } |
---|
| 908 | } |
---|
| 909 | |
---|
| 910 | |
---|
| 911 | /**Function******************************************************************** |
---|
| 912 | |
---|
| 913 | Synopsis [required] |
---|
| 914 | |
---|
| 915 | Description [optional] |
---|
| 916 | |
---|
| 917 | SideEffects [required] |
---|
| 918 | |
---|
| 919 | SeeAlso [optional] |
---|
| 920 | |
---|
| 921 | ******************************************************************************/ |
---|
| 922 | static void |
---|
| 923 | heur_touati_la2( |
---|
| 924 | Proc_Com_Graph * PCG, |
---|
| 925 | int * opt_perm) |
---|
| 926 | { |
---|
| 927 | int i; |
---|
| 928 | |
---|
| 929 | for(i=0; i<PCG->num_proc; i++) { |
---|
| 930 | opt_perm[i] = i; |
---|
| 931 | } |
---|
| 932 | |
---|
| 933 | for(i=0; i<(2*floor(((double)PCG->num_proc) / 2.0 )); i = i+2) { |
---|
| 934 | /*printf("Iteration\t:\t%d(%d)\n", i, PCG->num_proc );*/ |
---|
| 935 | append_touati_2(opt_perm,i,PCG); |
---|
| 936 | } |
---|
| 937 | } |
---|
| 938 | |
---|
| 939 | |
---|
| 940 | /**Function******************************************************************** |
---|
| 941 | |
---|
| 942 | Synopsis [required] |
---|
| 943 | |
---|
| 944 | Description [Returns the ordering of processes that minimizes the bound.] |
---|
| 945 | |
---|
| 946 | SideEffects [required] |
---|
| 947 | |
---|
| 948 | SeeAlso [optional] |
---|
| 949 | |
---|
| 950 | ******************************************************************************/ |
---|
| 951 | static int * |
---|
| 952 | opt_proc_order( |
---|
| 953 | Proc_Com_Graph * PCG) |
---|
| 954 | { |
---|
| 955 | int *cur_opt; |
---|
| 956 | |
---|
| 957 | |
---|
| 958 | cur_opt = ALLOC(int, PCG->num_proc); |
---|
| 959 | heur_2(PCG, cur_opt); |
---|
| 960 | |
---|
| 961 | /*for (i=0;i<PCG->num_proc; i++) printf(" %d ",cur_opt[i]+1);*/ |
---|
| 962 | |
---|
| 963 | return cur_opt; |
---|
| 964 | } |
---|
| 965 | |
---|
| 966 | |
---|
| 967 | /**Function******************************************************************** |
---|
| 968 | |
---|
| 969 | Synopsis [required] |
---|
| 970 | |
---|
| 971 | Description [Returns the ordering of processes that minimizes the bound.] |
---|
| 972 | |
---|
| 973 | SideEffects [required] |
---|
| 974 | |
---|
| 975 | SeeAlso [optional] |
---|
| 976 | |
---|
| 977 | ******************************************************************************/ |
---|
| 978 | static int * |
---|
| 979 | opt_touati_order( |
---|
| 980 | Proc_Com_Graph * PCG) |
---|
| 981 | { |
---|
| 982 | int *cur_opt; |
---|
| 983 | |
---|
| 984 | |
---|
| 985 | /* |
---|
| 986 | * FIX: same function as opt_proc_order except for heuristic call. |
---|
| 987 | */ |
---|
| 988 | |
---|
| 989 | cur_opt = ALLOC(int, PCG->num_proc); |
---|
| 990 | heur_touati_la2(PCG, cur_opt); |
---|
| 991 | |
---|
| 992 | /*for (i=0;i<PCG->num_proc; i++) printf(" %d ",cur_opt[i]+1);*/ |
---|
| 993 | |
---|
| 994 | return cur_opt; |
---|
| 995 | } |
---|
| 996 | |
---|
| 997 | /**Function******************************************************************** |
---|
| 998 | |
---|
| 999 | Synopsis [required] |
---|
| 1000 | |
---|
| 1001 | Description [optional] |
---|
| 1002 | |
---|
| 1003 | SideEffects [required] |
---|
| 1004 | |
---|
| 1005 | SeeAlso [optional] |
---|
| 1006 | |
---|
| 1007 | ******************************************************************************/ |
---|
| 1008 | static double |
---|
| 1009 | cost_total( |
---|
| 1010 | Proc_Com_Graph * PCG, |
---|
| 1011 | int * perm) |
---|
| 1012 | { |
---|
| 1013 | int i; |
---|
| 1014 | |
---|
| 1015 | double cost=0; |
---|
| 1016 | |
---|
| 1017 | for(i=0; i<PCG->num_proc-1; i++) { |
---|
| 1018 | cost += cost_for_cut(PCG, perm, i); |
---|
| 1019 | } |
---|
| 1020 | |
---|
| 1021 | return cost; |
---|
| 1022 | } |
---|
| 1023 | |
---|
| 1024 | |
---|
| 1025 | /**Function******************************************************************** |
---|
| 1026 | |
---|
| 1027 | Synopsis [required] |
---|
| 1028 | |
---|
| 1029 | Description [optional] |
---|
| 1030 | |
---|
| 1031 | SideEffects [required] |
---|
| 1032 | |
---|
| 1033 | SeeAlso [optional] |
---|
| 1034 | |
---|
| 1035 | ******************************************************************************/ |
---|
| 1036 | static int * |
---|
| 1037 | random_permutation( |
---|
| 1038 | int seed, |
---|
| 1039 | int n_elts) |
---|
| 1040 | { |
---|
| 1041 | int i, j; |
---|
| 1042 | int *permutation; |
---|
| 1043 | int *remaining; |
---|
| 1044 | int next_entry; |
---|
| 1045 | int next_value; |
---|
| 1046 | int n_entries; |
---|
| 1047 | |
---|
| 1048 | if (n_elts <= 0) { |
---|
| 1049 | return NIL(int); |
---|
| 1050 | } |
---|
| 1051 | |
---|
| 1052 | n_entries = n_elts; |
---|
| 1053 | permutation = ALLOC(int, n_entries); |
---|
| 1054 | remaining = ALLOC(int, n_entries); |
---|
| 1055 | for (i = 0; i < n_entries; i++) { |
---|
| 1056 | remaining[i] = i; |
---|
| 1057 | } |
---|
| 1058 | |
---|
| 1059 | util_srandom((long) seed); |
---|
| 1060 | |
---|
| 1061 | next_entry = 0; |
---|
| 1062 | for (; n_entries > 0; n_entries--) { |
---|
| 1063 | next_value = util_random() % n_entries; |
---|
| 1064 | permutation[next_entry++] = remaining[next_value]; |
---|
| 1065 | for (j = next_value; j < n_entries - 1; j++) { |
---|
| 1066 | remaining[j] = remaining[j + 1]; |
---|
| 1067 | } |
---|
| 1068 | } |
---|
| 1069 | FREE(remaining); |
---|
| 1070 | return permutation; |
---|
| 1071 | } |
---|