use of com.tencent.angel.ml.math2.storage.LongDoubleVectorStorage in project angel by Tencent.
the class SimpleBinaryInAllExecutor method apply.
public static Vector apply(LongDoubleVector v1, LongLongVector v2, Binary op) {
if (v1.isSparse() && v2.isSparse()) {
if (op.isKeepStorage()) {
throw new AngelException("operation is not support!");
} else {
// multi-rehash
LongDoubleVectorStorage newStorage = v1.getStorage().emptySparse((int) (v1.getDim()));
LongDoubleVectorStorage v1Storage = v1.getStorage();
LongLongVectorStorage v2Storage = v2.getStorage();
for (int i = 0; i < v1.getDim(); i++) {
if (v1Storage.hasKey(i) && v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(v1.get(i), v2.get(i)));
} else if (v1Storage.hasKey(i) && !v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(v1.get(i), 0));
} else if (!v1Storage.hasKey(i) && v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(0, v2.get(i)));
} else {
newStorage.set(i, op.apply(0.0, 0));
}
}
v1.setStorage(newStorage);
}
} else if (v1.isSparse() && v2.isSorted()) {
if (op.isKeepStorage()) {
throw new AngelException("operation is not support!");
} else {
LongDoubleVectorStorage newStorage = v1.getStorage().emptySparse((int) (v1.getDim()));
LongDoubleVectorStorage v1Storage = v1.getStorage();
LongLongVectorStorage v2Storage = v2.getStorage();
for (int i = 0; i < v1.getDim(); i++) {
if (v1Storage.hasKey(i) && v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(v1.get(i), v2.get(i)));
} else if (v1Storage.hasKey(i) && !v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(v1.get(i), 0));
} else if (!v1Storage.hasKey(i) && v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(0, v2.get(i)));
} else {
newStorage.set(i, op.apply(0.0, 0));
}
}
v1.setStorage(newStorage);
}
} else if (v1.isSorted() && v2.isSparse()) {
if (op.isKeepStorage()) {
throw new AngelException("operation is not support!");
} else {
LongDoubleVectorStorage newStorage = new LongDoubleSparseVectorStorage(v1.getDim());
LongDoubleVectorStorage v1Storage = v1.getStorage();
LongLongVectorStorage v2Storage = v2.getStorage();
for (int i = 0; i < v1.getDim(); i++) {
if (v1Storage.hasKey(i) && v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(v1.get(i), v2.get(i)));
} else if (v1Storage.hasKey(i) && !v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(v1.get(i), 0));
} else if (!v1Storage.hasKey(i) && v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(0, v2.get(i)));
} else {
newStorage.set(i, op.apply(0.0, 0));
}
}
v1.setStorage(newStorage);
}
} else if (v1.isSorted() && v2.isSorted()) {
if (op.isKeepStorage()) {
throw new AngelException("operation is not support!");
} else {
LongDoubleVectorStorage newStorage = v1.getStorage().emptySorted((int) (v1.getDim()));
long[] resIndices = newStorage.getIndices();
double[] resValues = newStorage.getValues();
int v1Pointor = 0;
int v2Pointor = 0;
long size1 = v1.size();
long size2 = v2.size();
long[] v1Indices = v1.getStorage().getIndices();
double[] v1Values = v1.getStorage().getValues();
long[] v2Indices = v2.getStorage().getIndices();
long[] v2Values = v2.getStorage().getValues();
if (!op.isCompare()) {
for (int i = 0; i < resValues.length; i++) {
resValues[i] = Double.NaN;
}
}
int globalPointor = 0;
while (v1Pointor < size1 && v2Pointor < size2) {
if (v1Indices[v1Pointor] == v2Indices[v2Pointor]) {
resIndices[globalPointor] = v1Indices[v1Pointor];
resValues[globalPointor] = op.apply(v1Values[v1Pointor], v2Values[v2Pointor]);
v1Pointor++;
v2Pointor++;
globalPointor++;
} else if (v1Indices[v1Pointor] < v2Indices[v2Pointor]) {
resIndices[globalPointor] = v1Indices[v1Pointor];
resValues[globalPointor] = op.apply(v2Values[v2Pointor], 0);
v1Pointor++;
globalPointor++;
} else {
// v1Indices[v1Pointor] > v2Indices[v2Pointor]
resIndices[globalPointor] = v2Indices[v2Pointor];
resValues[globalPointor] = op.apply(0, v2Values[v2Pointor]);
v2Pointor++;
globalPointor++;
}
}
v1.setStorage(newStorage);
}
} else {
throw new AngelException("The operation is not support!");
}
return v1;
}
use of com.tencent.angel.ml.math2.storage.LongDoubleVectorStorage in project angel by Tencent.
the class MixedBinaryOutAllExecutor method apply.
private static Vector apply(CompLongDoubleVector v1, LongLongVector v2, Binary op) {
LongDoubleVector[] parts = v1.getPartitions();
Storage[] resParts = StorageSwitch.applyComp(v1, v2, op);
if (!op.isKeepStorage()) {
for (int i = 0; i < parts.length; i++) {
if (parts[i].getStorage() instanceof LongDoubleSortedVectorStorage) {
resParts[i] = new LongDoubleSparseVectorStorage(parts[i].getDim(), parts[i].getStorage().getIndices(), parts[i].getStorage().getValues());
}
}
}
long subDim = (v1.getDim() + v1.getNumPartitions() - 1) / v1.getNumPartitions();
for (int i = 0; i < v1.getDim(); i++) {
int pidx = (int) (i / subDim);
long subidx = i % subDim;
if (v2.getStorage().hasKey(i)) {
((LongDoubleVectorStorage) resParts[pidx]).set(subidx, op.apply(parts[pidx].get(subidx), v2.get(i)));
} else {
((LongDoubleVectorStorage) resParts[pidx]).set(subidx, op.apply(parts[pidx].get(subidx), 0));
}
}
LongDoubleVector[] res = new LongDoubleVector[parts.length];
int i = 0;
for (LongDoubleVector part : parts) {
res[i] = new LongDoubleVector(part.getMatrixId(), part.getRowId(), part.getClock(), part.getDim(), (LongDoubleVectorStorage) resParts[i]);
i++;
}
return new CompLongDoubleVector(v1.getMatrixId(), v1.getRowId(), v1.getClock(), v1.getDim(), res, v1.getSubDim());
}
use of com.tencent.angel.ml.math2.storage.LongDoubleVectorStorage in project angel by Tencent.
the class SimpleBinaryOutNonZAExecutor method apply.
public static Vector apply(LongDoubleVector v1, LongDoubleVector v2, Binary op) {
LongDoubleVectorStorage newStorage = (LongDoubleVectorStorage) StorageSwitch.apply(v1, v2, op);
if (v1.isSparse() && v2.isSparse()) {
long v1Size = v1.size();
long v2Size = v2.size();
if (v1Size >= v2Size * Constant.sparseThreshold && (v1Size + v2Size) * Constant.intersectionCoeff <= Constant.sparseDenseStorageThreshold * v1.dim()) {
// we gauss the indices of v2 maybe is a subset of v1, or overlap is very large
ObjectIterator<Long2DoubleMap.Entry> iter = v2.getStorage().entryIterator();
while (iter.hasNext()) {
Long2DoubleMap.Entry entry = iter.next();
long idx = entry.getLongKey();
newStorage.set(idx, op.apply(v1.get(idx), entry.getDoubleValue()));
}
} else if ((v1Size + v2Size) * Constant.intersectionCoeff >= Constant.sparseDenseStorageThreshold * v1.dim()) {
// we gauss dense storage is more efficient
ObjectIterator<Long2DoubleMap.Entry> iter1 = v1.getStorage().entryIterator();
while (iter1.hasNext()) {
Long2DoubleMap.Entry entry = iter1.next();
long idx = entry.getLongKey();
newStorage.set(idx, entry.getDoubleValue());
}
ObjectIterator<Long2DoubleMap.Entry> iter2 = v2.getStorage().entryIterator();
while (iter2.hasNext()) {
Long2DoubleMap.Entry entry = iter2.next();
long idx = entry.getLongKey();
newStorage.set(idx, op.apply(v1.get(idx), entry.getDoubleValue()));
}
} else {
// to avoid multi-rehash
int capacity = 1 << (32 - Integer.numberOfLeadingZeros((int) (v1.size() / 0.75)));
if (v1.size() + v2.size() <= 1.5 * capacity) {
// no rehashor one onle rehash is required, nothing to optimization
ObjectIterator<Long2DoubleMap.Entry> iter = v2.getStorage().entryIterator();
while (iter.hasNext()) {
Long2DoubleMap.Entry entry = iter.next();
long idx = entry.getLongKey();
newStorage.set(idx, op.apply(v1.get(idx), entry.getDoubleValue()));
}
} else {
// multi-rehash
ObjectIterator<Long2DoubleMap.Entry> iter1 = v1.getStorage().entryIterator();
while (iter1.hasNext()) {
Long2DoubleMap.Entry entry = iter1.next();
long idx = entry.getLongKey();
newStorage.set(idx, entry.getDoubleValue());
}
ObjectIterator<Long2DoubleMap.Entry> iter2 = v2.getStorage().entryIterator();
while (iter2.hasNext()) {
Long2DoubleMap.Entry entry = iter2.next();
long idx = entry.getLongKey();
newStorage.set(idx, op.apply(v1.get(idx), entry.getDoubleValue()));
}
}
}
} else if (v1.isSparse() && v2.isSorted()) {
long v1Size = v1.size();
long v2Size = v2.size();
if (v1Size >= v2Size * Constant.sparseThreshold && (v1Size + v2Size) * Constant.intersectionCoeff <= Constant.sparseDenseStorageThreshold * v1.dim()) {
// we gauss the indices of v2 maybe is a subset of v1, or overlap is very large
long[] v2Indices = v2.getStorage().getIndices();
double[] v2Values = v2.getStorage().getValues();
for (int i = 0; i < v2.size(); i++) {
long idx = v2Indices[i];
newStorage.set(idx, op.apply(v1.get(idx), v2Values[i]));
}
} else if ((v1Size + v2Size) * Constant.intersectionCoeff >= Constant.sparseDenseStorageThreshold * v1.dim()) {
ObjectIterator<Long2DoubleMap.Entry> iter1 = v1.getStorage().entryIterator();
while (iter1.hasNext()) {
Long2DoubleMap.Entry entry = iter1.next();
long idx = entry.getLongKey();
newStorage.set(idx, entry.getDoubleValue());
}
long[] v2Indices = v2.getStorage().getIndices();
double[] v2Values = v2.getStorage().getValues();
long size = v2.size();
for (int i = 0; i < size; i++) {
long idx = v2Indices[i];
newStorage.set(idx, op.apply(v1.get(idx), v2Values[i]));
}
} else {
// to avoid multi-rehash
int capacity = 1 << (32 - Integer.numberOfLeadingZeros((int) (v1.size() / 0.75)));
if (v1.size() + v2.size() <= 1.5 * capacity) {
long[] v2Indices = v2.getStorage().getIndices();
double[] v2Values = v2.getStorage().getValues();
for (int i = 0; i < v2.size(); i++) {
long idx = v2Indices[i];
newStorage.set(idx, op.apply(v1.get(idx), v2Values[i]));
}
} else {
ObjectIterator<Long2DoubleMap.Entry> iter1 = v1.getStorage().entryIterator();
while (iter1.hasNext()) {
Long2DoubleMap.Entry entry = iter1.next();
long idx = entry.getLongKey();
newStorage.set(idx, entry.getDoubleValue());
}
long[] v2Indices = v2.getStorage().getIndices();
double[] v2Values = v2.getStorage().getValues();
long size = v2.size();
for (int i = 0; i < size; i++) {
long idx = v2Indices[i];
newStorage.set(idx, op.apply(v1.get(idx), v2Values[i]));
}
}
}
} else if (v1.isSorted() && v2.isSparse()) {
long v1Size = v1.size();
long v2Size = v2.size();
if ((v1Size + v2Size) * Constant.intersectionCoeff >= Constant.sortedDenseStorageThreshold * v1.dim()) {
if (op.isKeepStorage()) {
long[] v1Indices = v1.getStorage().getIndices();
long[] idxiter = v2.getStorage().indexIterator().toLongArray();
long[] indices = new long[(int) (v1Size + v2Size)];
System.arraycopy(v1Indices, 0, indices, 0, (int) v1.size());
System.arraycopy(idxiter, 0, indices, (int) v1.size(), (int) v2.size());
LongAVLTreeSet avl = new LongAVLTreeSet(indices);
LongBidirectionalIterator iter = avl.iterator();
double[] values = new double[indices.length];
int i = 0;
while (iter.hasNext()) {
long idx = iter.nextLong();
indices[i] = idx;
values[i] = op.apply(v1.get(idx), v2.get(idx));
i++;
}
while (i < indices.length) {
indices[i] = 0;
i++;
}
newStorage = new LongDoubleSortedVectorStorage(v1.getDim(), (int) avl.size(), indices, values);
} else {
long[] v1Indices = v1.getStorage().getIndices();
double[] v1Values = v1.getStorage().getValues();
long size = v1.size();
for (int i = 0; i < size; i++) {
long idx = v1Indices[i];
newStorage.set(idx, v1Values[i]);
}
ObjectIterator<Long2DoubleMap.Entry> iter = v2.getStorage().entryIterator();
while (iter.hasNext()) {
Long2DoubleMap.Entry entry = iter.next();
long idx = entry.getLongKey();
newStorage.set(idx, op.apply(newStorage.get(idx), entry.getDoubleValue()));
}
}
} else {
if (op.isKeepStorage()) {
long[] v1Indices = v1.getStorage().getIndices();
long[] idxiter = v2.getStorage().indexIterator().toLongArray();
long[] indices = new long[(int) (v1Size + v2Size)];
System.arraycopy(v1Indices, 0, indices, 0, (int) v1.size());
System.arraycopy(idxiter, 0, indices, (int) v1.size(), (int) v2.size());
LongAVLTreeSet avl = new LongAVLTreeSet(indices);
LongBidirectionalIterator iter = avl.iterator();
double[] values = new double[indices.length];
int i = 0;
while (iter.hasNext()) {
long idx = iter.nextLong();
indices[i] = idx;
values[i] = op.apply(v1.get(idx), v2.get(idx));
i++;
}
while (i < indices.length) {
indices[i] = 0;
i++;
}
newStorage = new LongDoubleSortedVectorStorage(v1.getDim(), (int) avl.size(), indices, values);
} else {
long[] v1Indices = v1.getStorage().getIndices();
double[] v1Values = v1.getStorage().getValues();
long size = v1.size();
for (int i = 0; i < size; i++) {
long idx = v1Indices[i];
newStorage.set(idx, v1Values[i]);
}
ObjectIterator<Long2DoubleMap.Entry> iter = v2.getStorage().entryIterator();
while (iter.hasNext()) {
Long2DoubleMap.Entry entry = iter.next();
long idx = entry.getLongKey();
newStorage.set(idx, op.apply(newStorage.get(idx), entry.getDoubleValue()));
}
}
}
} else if (v1.isSorted() && v2.isSorted()) {
int v1Pointor = 0;
int v2Pointor = 0;
long size1 = v1.size();
long size2 = v2.size();
long[] v1Indices = v1.getStorage().getIndices();
double[] v1Values = v1.getStorage().getValues();
long[] v2Indices = v2.getStorage().getIndices();
double[] v2Values = v2.getStorage().getValues();
if ((size1 + size2) * Constant.intersectionCoeff >= Constant.sortedDenseStorageThreshold * v1.dim()) {
if (op.isKeepStorage()) {
// sorted
long[] resIndices = newStorage.getIndices();
double[] resValues = newStorage.getValues();
int global = 0;
while (v1Pointor < size1 && v2Pointor < size2) {
if (v1Indices[v1Pointor] == v2Indices[v2Pointor]) {
resIndices[global] = v1Indices[v1Pointor];
resValues[global] = op.apply(v1Values[v1Pointor], v2Values[v2Pointor]);
global++;
v1Pointor++;
v2Pointor++;
} else if (v1Indices[v1Pointor] < v2Indices[v2Pointor]) {
resIndices[global] = v1Indices[v1Pointor];
resValues[global] = v1Values[v1Pointor];
global++;
v1Pointor++;
} else {
// v1Indices[v1Pointor] > v2Indices[v2Pointor]
resIndices[global] = v2Indices[v2Pointor];
resValues[global] = op.apply(0, v2Values[v2Pointor]);
global++;
v2Pointor++;
}
}
} else {
// dense
while (v1Pointor < size1 || v2Pointor < size2) {
if ((v1Pointor < size1 && v2Pointor < size2) && v1Indices[v1Pointor] == v2Indices[v2Pointor]) {
newStorage.set(v1Indices[v1Pointor], op.apply(v1Values[v1Pointor], v2Values[v2Pointor]));
v1Pointor++;
v2Pointor++;
} else if ((v1Pointor < size1 && v2Pointor < size2) && v1Indices[v1Pointor] < v2Indices[v2Pointor] || (v1Pointor < size1 && v2Pointor >= size2)) {
newStorage.set(v1Indices[v1Pointor], v1Values[v1Pointor]);
v1Pointor++;
} else if (((v1Pointor < size1 && v2Pointor < size2) && v1Indices[v1Pointor] >= v2Indices[v2Pointor]) || (v1Pointor >= size1 && v2Pointor < size2)) {
newStorage.set(v2Indices[v2Pointor], op.apply(0, v2Values[v2Pointor]));
v2Pointor++;
}
}
}
} else {
if (op.isKeepStorage()) {
long[] resIndices = newStorage.getIndices();
double[] resValues = newStorage.getValues();
int globalPointor = 0;
while (v1Pointor < size1 && v2Pointor < size2) {
if (v1Indices[v1Pointor] == v2Indices[v2Pointor]) {
resIndices[globalPointor] = v1Indices[v1Pointor];
resValues[globalPointor] = op.apply(v1Values[v1Pointor], v2Values[v2Pointor]);
v1Pointor++;
v2Pointor++;
globalPointor++;
} else if (v1Indices[v1Pointor] < v2Indices[v2Pointor]) {
resIndices[globalPointor] = v1Indices[v1Pointor];
resValues[globalPointor] = v1Values[v1Pointor];
v1Pointor++;
globalPointor++;
} else {
// v1Indices[v1Pointor] > v2Indices[v2Pointor]
resIndices[globalPointor] = v2Indices[v2Pointor];
resValues[globalPointor] = op.apply(0, v2Values[v2Pointor]);
v2Pointor++;
globalPointor++;
}
}
} else {
while (v1Pointor < size1 || v2Pointor < size2) {
if ((v1Pointor < size1 && v2Pointor < size2) && v1Indices[v1Pointor] == v2Indices[v2Pointor]) {
newStorage.set(v1Indices[v1Pointor], op.apply(v1Values[v1Pointor], v2Values[v2Pointor]));
v1Pointor++;
v2Pointor++;
} else if ((v1Pointor < size1 && v2Pointor < size2) && v1Indices[v1Pointor] < v2Indices[v2Pointor] || (v1Pointor < size1 && v2Pointor >= size2)) {
newStorage.set(v1Indices[v1Pointor], v1Values[v1Pointor]);
v1Pointor++;
} else if (((v1Pointor < size1 && v2Pointor < size2) && v1Indices[v1Pointor] >= v2Indices[v2Pointor]) || (v1Pointor >= size1 && v2Pointor < size2)) {
newStorage.set(v2Indices[v2Pointor], op.apply(0, v2Values[v2Pointor]));
v2Pointor++;
}
}
}
}
} else {
throw new AngelException("The operation is not support!");
}
return new LongDoubleVector(v1.getMatrixId(), v1.getRowId(), v1.getClock(), v1.getDim(), newStorage);
}
use of com.tencent.angel.ml.math2.storage.LongDoubleVectorStorage in project angel by Tencent.
the class SimpleBinaryOutNonZAExecutor method apply.
public static Vector apply(LongDoubleVector v1, LongLongVector v2, Binary op) {
LongDoubleVectorStorage newStorage = (LongDoubleVectorStorage) StorageSwitch.apply(v1, v2, op);
if (v1.isSparse() && v2.isSparse()) {
long v1Size = v1.size();
long v2Size = v2.size();
if (v1Size >= v2Size * Constant.sparseThreshold && (v1Size + v2Size) * Constant.intersectionCoeff <= Constant.sparseDenseStorageThreshold * v1.dim()) {
// we gauss the indices of v2 maybe is a subset of v1, or overlap is very large
ObjectIterator<Long2LongMap.Entry> iter = v2.getStorage().entryIterator();
while (iter.hasNext()) {
Long2LongMap.Entry entry = iter.next();
long idx = entry.getLongKey();
newStorage.set(idx, op.apply(v1.get(idx), entry.getLongValue()));
}
} else if ((v1Size + v2Size) * Constant.intersectionCoeff >= Constant.sparseDenseStorageThreshold * v1.dim()) {
// we gauss dense storage is more efficient
ObjectIterator<Long2DoubleMap.Entry> iter1 = v1.getStorage().entryIterator();
while (iter1.hasNext()) {
Long2DoubleMap.Entry entry = iter1.next();
long idx = entry.getLongKey();
newStorage.set(idx, entry.getDoubleValue());
}
ObjectIterator<Long2LongMap.Entry> iter2 = v2.getStorage().entryIterator();
while (iter2.hasNext()) {
Long2LongMap.Entry entry = iter2.next();
long idx = entry.getLongKey();
newStorage.set(idx, op.apply(v1.get(idx), entry.getLongValue()));
}
} else {
// to avoid multi-rehash
int capacity = 1 << (32 - Integer.numberOfLeadingZeros((int) (v1.size() / 0.75)));
if (v1.size() + v2.size() <= 1.5 * capacity) {
// no rehashor one onle rehash is required, nothing to optimization
ObjectIterator<Long2LongMap.Entry> iter = v2.getStorage().entryIterator();
while (iter.hasNext()) {
Long2LongMap.Entry entry = iter.next();
long idx = entry.getLongKey();
newStorage.set(idx, op.apply(v1.get(idx), entry.getLongValue()));
}
} else {
// multi-rehash
ObjectIterator<Long2DoubleMap.Entry> iter1 = v1.getStorage().entryIterator();
while (iter1.hasNext()) {
Long2DoubleMap.Entry entry = iter1.next();
long idx = entry.getLongKey();
newStorage.set(idx, entry.getDoubleValue());
}
ObjectIterator<Long2LongMap.Entry> iter2 = v2.getStorage().entryIterator();
while (iter2.hasNext()) {
Long2LongMap.Entry entry = iter2.next();
long idx = entry.getLongKey();
newStorage.set(idx, op.apply(v1.get(idx), entry.getLongValue()));
}
}
}
} else if (v1.isSparse() && v2.isSorted()) {
long v1Size = v1.size();
long v2Size = v2.size();
if (v1Size >= v2Size * Constant.sparseThreshold && (v1Size + v2Size) * Constant.intersectionCoeff <= Constant.sparseDenseStorageThreshold * v1.dim()) {
// we gauss the indices of v2 maybe is a subset of v1, or overlap is very large
long[] v2Indices = v2.getStorage().getIndices();
long[] v2Values = v2.getStorage().getValues();
for (int i = 0; i < v2.size(); i++) {
long idx = v2Indices[i];
newStorage.set(idx, op.apply(v1.get(idx), v2Values[i]));
}
} else if ((v1Size + v2Size) * Constant.intersectionCoeff >= Constant.sparseDenseStorageThreshold * v1.dim()) {
ObjectIterator<Long2DoubleMap.Entry> iter1 = v1.getStorage().entryIterator();
while (iter1.hasNext()) {
Long2DoubleMap.Entry entry = iter1.next();
long idx = entry.getLongKey();
newStorage.set(idx, entry.getDoubleValue());
}
long[] v2Indices = v2.getStorage().getIndices();
long[] v2Values = v2.getStorage().getValues();
long size = v2.size();
for (int i = 0; i < size; i++) {
long idx = v2Indices[i];
newStorage.set(idx, op.apply(v1.get(idx), v2Values[i]));
}
} else {
// to avoid multi-rehash
int capacity = 1 << (32 - Integer.numberOfLeadingZeros((int) (v1.size() / 0.75)));
if (v1.size() + v2.size() <= 1.5 * capacity) {
long[] v2Indices = v2.getStorage().getIndices();
long[] v2Values = v2.getStorage().getValues();
for (int i = 0; i < v2.size(); i++) {
long idx = v2Indices[i];
newStorage.set(idx, op.apply(v1.get(idx), v2Values[i]));
}
} else {
ObjectIterator<Long2DoubleMap.Entry> iter1 = v1.getStorage().entryIterator();
while (iter1.hasNext()) {
Long2DoubleMap.Entry entry = iter1.next();
long idx = entry.getLongKey();
newStorage.set(idx, entry.getDoubleValue());
}
long[] v2Indices = v2.getStorage().getIndices();
long[] v2Values = v2.getStorage().getValues();
long size = v2.size();
for (int i = 0; i < size; i++) {
long idx = v2Indices[i];
newStorage.set(idx, op.apply(v1.get(idx), v2Values[i]));
}
}
}
} else if (v1.isSorted() && v2.isSparse()) {
long v1Size = v1.size();
long v2Size = v2.size();
if ((v1Size + v2Size) * Constant.intersectionCoeff >= Constant.sortedDenseStorageThreshold * v1.dim()) {
if (op.isKeepStorage()) {
long[] v1Indices = v1.getStorage().getIndices();
long[] idxiter = v2.getStorage().indexIterator().toLongArray();
long[] indices = new long[(int) (v1Size + v2Size)];
System.arraycopy(v1Indices, 0, indices, 0, (int) v1.size());
System.arraycopy(idxiter, 0, indices, (int) v1.size(), (int) v2.size());
LongAVLTreeSet avl = new LongAVLTreeSet(indices);
LongBidirectionalIterator iter = avl.iterator();
double[] values = new double[indices.length];
int i = 0;
while (iter.hasNext()) {
long idx = iter.nextLong();
indices[i] = idx;
values[i] = op.apply(v1.get(idx), v2.get(idx));
i++;
}
while (i < indices.length) {
indices[i] = 0;
i++;
}
newStorage = new LongDoubleSortedVectorStorage(v1.getDim(), (int) avl.size(), indices, values);
} else {
long[] v1Indices = v1.getStorage().getIndices();
double[] v1Values = v1.getStorage().getValues();
long size = v1.size();
for (int i = 0; i < size; i++) {
long idx = v1Indices[i];
newStorage.set(idx, v1Values[i]);
}
ObjectIterator<Long2LongMap.Entry> iter = v2.getStorage().entryIterator();
while (iter.hasNext()) {
Long2LongMap.Entry entry = iter.next();
long idx = entry.getLongKey();
newStorage.set(idx, op.apply(newStorage.get(idx), entry.getLongValue()));
}
}
} else {
if (op.isKeepStorage()) {
long[] v1Indices = v1.getStorage().getIndices();
long[] idxiter = v2.getStorage().indexIterator().toLongArray();
long[] indices = new long[(int) (v1Size + v2Size)];
System.arraycopy(v1Indices, 0, indices, 0, (int) v1.size());
System.arraycopy(idxiter, 0, indices, (int) v1.size(), (int) v2.size());
LongAVLTreeSet avl = new LongAVLTreeSet(indices);
LongBidirectionalIterator iter = avl.iterator();
double[] values = new double[indices.length];
int i = 0;
while (iter.hasNext()) {
long idx = iter.nextLong();
indices[i] = idx;
values[i] = op.apply(v1.get(idx), v2.get(idx));
i++;
}
while (i < indices.length) {
indices[i] = 0;
i++;
}
newStorage = new LongDoubleSortedVectorStorage(v1.getDim(), (int) avl.size(), indices, values);
} else {
long[] v1Indices = v1.getStorage().getIndices();
double[] v1Values = v1.getStorage().getValues();
long size = v1.size();
for (int i = 0; i < size; i++) {
long idx = v1Indices[i];
newStorage.set(idx, v1Values[i]);
}
ObjectIterator<Long2LongMap.Entry> iter = v2.getStorage().entryIterator();
while (iter.hasNext()) {
Long2LongMap.Entry entry = iter.next();
long idx = entry.getLongKey();
newStorage.set(idx, op.apply(newStorage.get(idx), entry.getLongValue()));
}
}
}
} else if (v1.isSorted() && v2.isSorted()) {
int v1Pointor = 0;
int v2Pointor = 0;
long size1 = v1.size();
long size2 = v2.size();
long[] v1Indices = v1.getStorage().getIndices();
double[] v1Values = v1.getStorage().getValues();
long[] v2Indices = v2.getStorage().getIndices();
long[] v2Values = v2.getStorage().getValues();
if ((size1 + size2) * Constant.intersectionCoeff >= Constant.sortedDenseStorageThreshold * v1.dim()) {
if (op.isKeepStorage()) {
// sorted
long[] resIndices = newStorage.getIndices();
double[] resValues = newStorage.getValues();
int global = 0;
while (v1Pointor < size1 && v2Pointor < size2) {
if (v1Indices[v1Pointor] == v2Indices[v2Pointor]) {
resIndices[global] = v1Indices[v1Pointor];
resValues[global] = op.apply(v1Values[v1Pointor], v2Values[v2Pointor]);
global++;
v1Pointor++;
v2Pointor++;
} else if (v1Indices[v1Pointor] < v2Indices[v2Pointor]) {
resIndices[global] = v1Indices[v1Pointor];
resValues[global] = v1Values[v1Pointor];
global++;
v1Pointor++;
} else {
// v1Indices[v1Pointor] > v2Indices[v2Pointor]
resIndices[global] = v2Indices[v2Pointor];
resValues[global] = op.apply(0, v2Values[v2Pointor]);
global++;
v2Pointor++;
}
}
} else {
// dense
while (v1Pointor < size1 || v2Pointor < size2) {
if ((v1Pointor < size1 && v2Pointor < size2) && v1Indices[v1Pointor] == v2Indices[v2Pointor]) {
newStorage.set(v1Indices[v1Pointor], op.apply(v1Values[v1Pointor], v2Values[v2Pointor]));
v1Pointor++;
v2Pointor++;
} else if ((v1Pointor < size1 && v2Pointor < size2) && v1Indices[v1Pointor] < v2Indices[v2Pointor] || (v1Pointor < size1 && v2Pointor >= size2)) {
newStorage.set(v1Indices[v1Pointor], v1Values[v1Pointor]);
v1Pointor++;
} else if (((v1Pointor < size1 && v2Pointor < size2) && v1Indices[v1Pointor] >= v2Indices[v2Pointor]) || (v1Pointor >= size1 && v2Pointor < size2)) {
newStorage.set(v2Indices[v2Pointor], op.apply(0, v2Values[v2Pointor]));
v2Pointor++;
}
}
}
} else {
if (op.isKeepStorage()) {
long[] resIndices = newStorage.getIndices();
double[] resValues = newStorage.getValues();
int globalPointor = 0;
while (v1Pointor < size1 && v2Pointor < size2) {
if (v1Indices[v1Pointor] == v2Indices[v2Pointor]) {
resIndices[globalPointor] = v1Indices[v1Pointor];
resValues[globalPointor] = op.apply(v1Values[v1Pointor], v2Values[v2Pointor]);
v1Pointor++;
v2Pointor++;
globalPointor++;
} else if (v1Indices[v1Pointor] < v2Indices[v2Pointor]) {
resIndices[globalPointor] = v1Indices[v1Pointor];
resValues[globalPointor] = v1Values[v1Pointor];
v1Pointor++;
globalPointor++;
} else {
// v1Indices[v1Pointor] > v2Indices[v2Pointor]
resIndices[globalPointor] = v2Indices[v2Pointor];
resValues[globalPointor] = op.apply(0, v2Values[v2Pointor]);
v2Pointor++;
globalPointor++;
}
}
} else {
while (v1Pointor < size1 || v2Pointor < size2) {
if ((v1Pointor < size1 && v2Pointor < size2) && v1Indices[v1Pointor] == v2Indices[v2Pointor]) {
newStorage.set(v1Indices[v1Pointor], op.apply(v1Values[v1Pointor], v2Values[v2Pointor]));
v1Pointor++;
v2Pointor++;
} else if ((v1Pointor < size1 && v2Pointor < size2) && v1Indices[v1Pointor] < v2Indices[v2Pointor] || (v1Pointor < size1 && v2Pointor >= size2)) {
newStorage.set(v1Indices[v1Pointor], v1Values[v1Pointor]);
v1Pointor++;
} else if (((v1Pointor < size1 && v2Pointor < size2) && v1Indices[v1Pointor] >= v2Indices[v2Pointor]) || (v1Pointor >= size1 && v2Pointor < size2)) {
newStorage.set(v2Indices[v2Pointor], op.apply(0, v2Values[v2Pointor]));
v2Pointor++;
}
}
}
}
} else {
throw new AngelException("The operation is not support!");
}
return new LongDoubleVector(v1.getMatrixId(), v1.getRowId(), v1.getClock(), v1.getDim(), newStorage);
}
use of com.tencent.angel.ml.math2.storage.LongDoubleVectorStorage in project angel by Tencent.
the class SimpleBinaryOutAllExecutor method apply.
public static Vector apply(LongDoubleVector v1, LongFloatVector v2, Binary op) {
LongDoubleVector res;
if (v1.isSparse() && v2.isSparse()) {
if (op.isKeepStorage()) {
throw new AngelException("operation is not support!");
} else {
// multi-rehash
LongDoubleVectorStorage newStorage = v1.getStorage().emptySparse((int) (v1.getDim()));
LongDoubleVectorStorage v1Storage = v1.getStorage();
LongFloatVectorStorage v2Storage = v2.getStorage();
for (int i = 0; i < v1.getDim(); i++) {
if (v1Storage.hasKey(i) && v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(v1.get(i), v2.get(i)));
} else if (v1Storage.hasKey(i) && !v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(v1.get(i), 0));
} else if (!v1Storage.hasKey(i) && v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(0, v2.get(i)));
} else {
newStorage.set(i, op.apply(0.0, 0));
}
}
res = new LongDoubleVector(v1.getMatrixId(), v1.getRowId(), v1.getClock(), v1.getDim(), newStorage);
}
} else if (v1.isSparse() && v2.isSorted()) {
if (op.isKeepStorage()) {
throw new AngelException("operation is not support!");
} else {
LongDoubleVectorStorage newStorage = v1.getStorage().emptySparse((int) (v1.getDim()));
LongDoubleVectorStorage v1Storage = v1.getStorage();
LongFloatVectorStorage v2Storage = v2.getStorage();
for (int i = 0; i < v1.getDim(); i++) {
if (v1Storage.hasKey(i) && v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(v1.get(i), v2.get(i)));
} else if (v1Storage.hasKey(i) && !v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(v1.get(i), 0));
} else if (!v1Storage.hasKey(i) && v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(0, v2.get(i)));
} else {
newStorage.set(i, op.apply(0.0, 0));
}
}
res = new LongDoubleVector(v1.getMatrixId(), v1.getRowId(), v1.getClock(), v1.getDim(), newStorage);
}
} else if (v1.isSorted() && v2.isSparse()) {
if (op.isKeepStorage()) {
throw new AngelException("operation is not support!");
} else {
LongDoubleVectorStorage newStorage = new LongDoubleSparseVectorStorage(v1.getDim());
LongDoubleVectorStorage v1Storage = v1.getStorage();
LongFloatVectorStorage v2Storage = v2.getStorage();
for (int i = 0; i < v1.getDim(); i++) {
if (v1Storage.hasKey(i) && v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(v1.get(i), v2.get(i)));
} else if (v1Storage.hasKey(i) && !v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(v1.get(i), 0));
} else if (!v1Storage.hasKey(i) && v2Storage.hasKey(i)) {
newStorage.set(i, op.apply(0, v2.get(i)));
} else {
newStorage.set(i, op.apply(0.0, 0));
}
}
res = new LongDoubleVector(v1.getMatrixId(), v1.getRowId(), v1.getClock(), v1.getDim(), newStorage);
}
} else if (v1.isSorted() && v2.isSorted()) {
if (op.isKeepStorage()) {
throw new AngelException("operation is not support!");
} else {
LongDoubleVectorStorage newStorage = v1.getStorage().emptySorted((int) (v1.getDim()));
long[] resIndices = newStorage.getIndices();
double[] resValues = newStorage.getValues();
int v1Pointor = 0;
int v2Pointor = 0;
long size1 = v1.size();
long size2 = v2.size();
long[] v1Indices = v1.getStorage().getIndices();
double[] v1Values = v1.getStorage().getValues();
long[] v2Indices = v2.getStorage().getIndices();
float[] v2Values = v2.getStorage().getValues();
if (!op.isCompare()) {
for (int i = 0; i < resValues.length; i++) {
resValues[i] = Double.NaN;
}
}
int globalPointor = 0;
while (v1Pointor < size1 && v2Pointor < size2) {
if (v1Indices[v1Pointor] == v2Indices[v2Pointor]) {
resIndices[globalPointor] = v1Indices[v1Pointor];
resValues[globalPointor] = op.apply(v1Values[v1Pointor], v2Values[v2Pointor]);
v1Pointor++;
v2Pointor++;
globalPointor++;
} else if (v1Indices[v1Pointor] < v2Indices[v2Pointor]) {
resIndices[globalPointor] = v1Indices[v1Pointor];
resValues[globalPointor] = op.apply(v2Values[v2Pointor], 0);
v1Pointor++;
globalPointor++;
} else {
// v1Indices[v1Pointor] > v2Indices[v2Pointor]
resIndices[globalPointor] = v2Indices[v2Pointor];
resValues[globalPointor] = op.apply(0, v2Values[v2Pointor]);
v2Pointor++;
globalPointor++;
}
}
res = new LongDoubleVector(v1.getMatrixId(), v1.getRowId(), v1.getClock(), v1.getDim(), newStorage);
}
} else {
throw new AngelException("The operation is not support!");
}
return res;
}
Aggregations