}
uint convertSize(uint cost) {
- cost = FUDGEFACTOR * cost;// fudge factor
return NEXTPOW2(cost);
}
EncodingSubGraph *tmpsg = leftGraph; leftGraph = rightGraph; rightGraph = tmpsg;
}
- uint leftSize = 0, rightSize = 0, newSize = 0, max = 0;
+ uint leftSize = 0, rightSize = 0, newSize = 0, min = 0;
bool merge = false;
if (leftGraph == NULL && rightGraph == NULL) {
leftSize = convertSize(left->getSize());
newSize = convertSize(left->s->getUnionSize(right->s));
newSize = (leftSize > newSize) ? leftSize : newSize;
newSize = (rightSize > newSize) ? rightSize : newSize;
- max = rightSize > leftSize ? rightSize : leftSize;
- merge = left->measureSimilarity(right) > 1.5 || max == newSize;
+ min = rightSize > leftSize ? leftSize : rightSize;
+ merge = left->measureSimilarity(right) > 1.5 || min == newSize;
} else if (leftGraph != NULL && rightGraph == NULL) {
- leftSize = convertSize(leftGraph->encodingSize);
+ leftSize = convertSize(leftGraph->numValues());
rightSize = convertSize(right->getSize());
newSize = convertSize(leftGraph->estimateNewSize(right));
newSize = (leftSize > newSize) ? leftSize : newSize;
newSize = (rightSize > newSize) ? rightSize : newSize;
- max = rightSize > leftSize ? rightSize : leftSize;
+ min = rightSize > leftSize ? leftSize : rightSize;
// model_print("Merge=%s\tsimilarity=%f\n", max==newSize?"TRUE":"FALSE", left->measureSimilarity(right));
- merge = left->measureSimilarity(right) > 1.5 || max == newSize;
+ merge = leftGraph->measureSimilarity(right) > 1.5 || min == newSize;
} else {
//Neither are null
- leftSize = convertSize(leftGraph->encodingSize);
- rightSize = convertSize(rightGraph->encodingSize);
+ leftSize = convertSize(leftGraph->numValues());
+ rightSize = convertSize(rightGraph->numValues());
newSize = convertSize(leftGraph->estimateNewSize(rightGraph));
// model_print("MergingSubGraphs: left=%u\tright=%u\tnewSize=%u\n", leftSize, rightSize, newSize);
newSize = (leftSize > newSize) ? leftSize : newSize;
newSize = (rightSize > newSize) ? rightSize : newSize;
- max = rightSize > leftSize ? rightSize : leftSize;
+ min = rightSize > leftSize ? leftSize : rightSize;
// model_print("Merge=%s\tsimilarity=%f\n", max==newSize?"TRUE":"FALSE", leftGraph->measureSimilarity(right));
- merge = leftGraph->measureSimilarity(right) > 1.5 || max == newSize;
+ merge = leftGraph->measureSimilarity(rightGraph) > 1.5 || min == newSize;
}
if (merge) {
//add the edge
return s->getType();
}
-bool EncodingNode::itemExists(uint64_t item) {
- for (uint i = 0; i < s->getSize(); i++) {
- if (item == s->getElement(i))
- return true;
- }
- return false;
-}
-
double EncodingNode::measureSimilarity(EncodingNode *node) {
uint common = 0;
- for (uint i = 0; i < s->getSize(); i++) {
+ for (uint i = 0, j = 0; i < s->getSize() && j < node->s->getSize(); ) {
uint64_t item = s->getElement(i);
- if (node->itemExists(item)) {
+ uint64_t item2 = node->s->getElement(j);
+ if (item < item2)
+ i++;
+ else if (item2 > item)
+ j++;
+ else {
+ i++;
+ j++;
common++;
}
}
-// model_print("common=%u\tsize1=%u\tsize2=%u\tsim1=%f\tsim2=%f\n", common, s->getSize(), node->getSize(), 1.0*common/s->getSize(), 1.0*common/node->getSize());
+
return common * 1.0 / s->getSize() + common * 1.0 / node->getSize();
}
#include "structs.h"
#include "graphstructs.h"
-#define FUDGEFACTOR 1.2
-#define CONVERSIONFACTOR 0.5
class EncodingGraph {
public:
double measureSimilarity(EncodingNode *node);
void setEncoding(ElementEncodingType e) {encoding = e;}
ElementEncodingType getEncoding() {return encoding;}
- bool itemExists(uint64_t item);
bool couldBeBinaryIndex() {return encoding == BINARYINDEX || encoding == ELEM_UNASSIGNED;}
CMEMALLOC;
private:
#include "qsort.h"
EncodingSubGraph::EncodingSubGraph() :
- encodingSize(0),
- numElements(0),
maxEncodingVal(0) {
}
}
uint EncodingSubGraph::estimateNewSize(EncodingSubGraph *sg) {
- uint newSize = 0;
- SetIteratorEncodingNode *nit = sg->nodes.iterator();
- while (nit->hasNext()) {
- EncodingNode *en = nit->next();
- uint size = estimateNewSize(en);
- if (size > newSize)
- newSize = size;
+ uint newSize = sg->allValues.getSize() + allValues.getSize();
+ SetIterator64Int *it = sg->allValues.iterator();
+
+ while (it->hasNext()) {
+ uint64_t val = it->next();
+ if (allValues.contains(val))
+ newSize--;
}
- delete nit;
+ delete it;
return newSize;
}
double EncodingSubGraph::measureSimilarity(EncodingNode *node) {
uint common = 0;
- Hashset64Int intSet;
- SetIteratorEncodingNode *nit = nodes.iterator();
- while (nit->hasNext()) {
- EncodingNode *en = nit->next();
- for (uint i = 0; i < en->getSize(); i++) {
- intSet.add(en->getIndex(i));
- }
- }
- for (uint i = 0; i < node->getSize(); i++) {
- if (intSet.contains( node->getIndex(i) )) {
+ uint size = node->getSize();
+ for (uint i = 0; i < size; i++) {
+ uint64_t val = node->getIndex(i);
+ if (allValues.contains(val))
common++;
- }
}
-// model_print("measureSimilarity:139: common=%u\t GraphSize=%u\tnodeSize=%u\tGraphSim=%f\tnodeSim=%f\n", common, intSet.getSize(), node->getSize(), 1.0*common/intSet.getSize(), 1.0*common/node->getSize());
- delete nit;
- return common * 1.0 / intSet.getSize() + common * 1.0 / node->getSize();
+ return common * 1.0 / allValues.getSize() + common * 1.0 / node->getSize();
}
double EncodingSubGraph::measureSimilarity(EncodingSubGraph *sg) {
uint common = 0;
- Hashset64Int set1;
- Hashset64Int set2;
- SetIteratorEncodingNode *nit = nodes.iterator();
- while (nit->hasNext()) {
- EncodingNode *en = nit->next();
- for (uint i = 0; i < en->getSize(); i++) {
- set1.add(en->getIndex(i));
- }
- }
- delete nit;
- nit = sg->nodes.iterator();
- while (nit->hasNext()) {
- EncodingNode *en = nit->next();
- for (uint i = 0; i < en->getSize(); i++) {
- set2.add(en->getIndex(i));
- }
- }
- delete nit;
- SetIterator64Int *setIter1 = set1.iterator();
+ SetIterator64Int *setIter1 = allValues.iterator();
while (setIter1->hasNext()) {
uint64_t item1 = setIter1->next();
- if ( set2.contains(item1)) {
+ if ( sg->allValues.contains(item1)) {
common++;
}
}
delete setIter1;
-// model_print("measureSimilarity:139: common=%u\tGraphSize1=%u\tGraphSize2=%u\tGraphSize1=%f\tGraphSize2=%f\n", common, set1.getSize(), set2.getSize(), 1.0*common/set1.getSize(), 1.0*common/set2.getSize());
- return common * 1.0 / set1.getSize() + common * 1.0 / set2.getSize();
+ return common * 1.0 / allValues.getSize() + common * 1.0 / sg->allValues.getSize();
}
uint EncodingSubGraph::estimateNewSize(EncodingNode *n) {
- SetIteratorEncodingEdge *eeit = n->edges.iterator();
- uint newsize = n->getSize();
- while (eeit->hasNext()) {
- EncodingEdge *ee = eeit->next();
- if (ee->left != NULL && ee->left != n && nodes.contains(ee->left)) {
- uint intersectSize = n->s->getUnionSize(ee->left->s);
- if (intersectSize > newsize)
- newsize = intersectSize;
- }
- if (ee->right != NULL && ee->right != n && nodes.contains(ee->right)) {
- uint intersectSize = n->s->getUnionSize(ee->right->s);
- if (intersectSize > newsize)
- newsize = intersectSize;
- }
- if (ee->dst != NULL && ee->dst != n && nodes.contains(ee->dst)) {
- uint intersectSize = n->s->getUnionSize(ee->dst->s);
- if (intersectSize > newsize)
- newsize = intersectSize;
- }
+ uint nSize = n->getSize();
+ uint newSize = allValues.getSize() + nSize;
+ for (uint i = 0; i < nSize; i++) {
+ if (allValues.contains(n->getIndex(i)))
+ newSize--;
}
- delete eeit;
- return newsize;
+ return newSize;
+}
+
+uint EncodingSubGraph::numValues() {
+ return allValues.getSize();
}
void EncodingSubGraph::addNode(EncodingNode *n) {
nodes.add(n);
- uint newSize = estimateNewSize(n);
- numElements += n->elements.getSize();
- if (newSize > encodingSize)
- encodingSize = newSize;
+ uint size = n->getSize();
+ for (uint i = 0; i < size; i++)
+ allValues.add(n->getIndex(i));
}
SetIteratorEncodingNode *EncodingSubGraph::nodeIterator() {
uint getEncodingMaxVal(EncodingNode *n) { return maxEncodingVal;}
double measureSimilarity(EncodingNode *n);
double measureSimilarity(EncodingSubGraph *sg);
+ uint numValues();
CMEMALLOC;
private:
uint estimateNewSize(EncodingNode *n);
HashsetEncodingValue values;
HashsetEncodingNode nodes;
NVPMap map;
- uint encodingSize;
- uint numElements;
uint maxEncodingVal;
-
+ Hashset64Int allValues;
friend class EncodingGraph;
};
if (encoding->element->anyValue) {
uint setSize = encoding->element->getRange()->getSize();
int maxIndex = getMaximumUsedIndex(encoding);
- if (maxIndex !=-1 && (maxIndex - setSize) != 0 && (setSize/(maxIndex-setSize)) < pow(1.9, (uint)solver->getTuner()->getTunable(MUSTVALUE, &mustValueBinaryIndex) - 4)) {
+ if (maxIndex != -1 && (maxIndex - setSize) != 0 && (setSize / (maxIndex - setSize)) < pow(1.9, (uint)solver->getTuner()->getTunable(MUSTVALUE, &mustValueBinaryIndex) - 4)) {
generateAnyValueBinaryIndexEncodingPositive(encoding);
} else {
generateAnyValueBinaryIndexEncoding(encoding);
}
}
-int SATEncoder::getMaximumUsedIndex(ElementEncoding *encoding){
+int SATEncoder::getMaximumUsedIndex(ElementEncoding *encoding) {
int index = -1;
for (uint i = encoding->encArraySize - 1; i >= 0; i--) {
if (encoding->isinUseElement(i)) {
}
void MultiTuner::findBestThreeTuners() {
- if(allTuners.getSize() < 3){
+ if (allTuners.getSize() < 3) {
printData();
return;
}
- TunerRecord * bestTuners[3];
+ TunerRecord *bestTuners[3];
double score = DBL_MAX;
- for (uint i = 0; i < allTuners.getSize()-2; i++) {
- for(uint j=i+1; j< allTuners.getSize()-1; j++){
- for(uint k=j+1; k<allTuners.getSize(); k++){
+ for (uint i = 0; i < allTuners.getSize() - 2; i++) {
+ for (uint j = i + 1; j < allTuners.getSize() - 1; j++) {
+ for (uint k = j + 1; k < allTuners.getSize(); k++) {
TunerRecord *tuner1 = allTuners.get(i);
TunerRecord *tuner2 = allTuners.get(j);
TunerRecord *tuner3 = allTuners.get(k);
double mintimes[problems.getSize()];
- for(uint l=0; l < problems.getSize(); l++){
+ for (uint l = 0; l < problems.getSize(); l++) {
Problem *problem = problems.get(l);
mintimes[l] = pow(min(tuner1->getTime(problem), tuner2->getTime(problem), tuner3->getTime(problem)), (double)1 / problems.getSize());
}
- double result=1;
- for(uint l=0; l< problems.getSize(); l++){
+ double result = 1;
+ for (uint l = 0; l < problems.getSize(); l++) {
result *= mintimes[l];
}
- if(result < score){
+ if (result < score) {
score = result;
bestTuners[0] = tuner1;
bestTuners[1] = tuner2;
}
}
model_print("Best 3 tuners:\n");
- for(uint i=0; i<3; i++){
- TunerRecord * tuner = bestTuners[i];
+ for (uint i = 0; i < 3; i++) {
+ TunerRecord *tuner = bestTuners[i];
SearchTuner *stun = tuner->getTuner();
model_print("Tuner %u\n", tuner->tunernumber);
stun->print();
void mapProblemsToTuners(Vector<TunerRecord *> *tunerV);
void improveTuners(Vector<TunerRecord *> *tunerV);
TunerRecord *tune(TunerRecord *tuner);
- inline long long min(long long num1, long long num2, long long num3){
- return num1 < num2 && num1 < num3? num1:
- num2 < num3? num2 : num3;
- }
-
Vector<TunerRecord *> allTuners;
Vector<Problem *> problems;
Vector<TunerRecord *> tuners;
uint timeout;
int execnum;
};
+
+inline long long min(long long num1, long long num2, long long num3) {
+ return num1 < num2 && num1 < num3 ? num1 :
+ num2 < num3 ? num2 : num3;
+}
+
#endif
projects / satune.git / commitdiff