inline const typename A::BASE_TYPE Sum(const VectorExpression<T2, A> &expr, bool concurrent = false) {
typename A::BASE_TYPE ret; // = TT(0.0);
if (concurrent) {
int range = expr.Size(0) / std::thread::hardware_concurrency();
std::vector<std::thread> pool;
std::vector<typename A::BASE_TYPE > temp(std::thread::hardware_concurrency());
for (int i = 0; i < std::thread::hardware_concurrency(); i++) {
if (i < (std::thread::hardware_concurrency() - 1)) {
// std::cout << i*range << " - " << (i + 1) * range << "\n";
pool.push_back(std::thread(SumThread<T2, A>, std::ref(temp[i]), i*range, (i + 1) * range, std::ref(expr)));
} else {
pool.push_back(std::thread(SumThread<T2, A>, std::ref(temp[i]), i*range, expr.Size(0), std::ref(expr)));
// std::cout << i*range << " - " << expr.Size(0) << "\n";
}
}
for (int i = 0; i < pool.size(); i++) {
pool[i].join();
}
for (int i = 0; i < std::thread::hardware_concurrency(); i++) {
ret += temp[i];
}
} else {
size_t s = expr.Size(0);
size_t end = (((s - 1UL) & size_t(-2)) + 1UL);
ret = expr(0);
for (size_t i = 1UL; i < end; i += 2UL) {
ret += expr(i) + expr(i + 1);
}
if (end < s) {
ret += expr(end);
}
}
return ret;
}
Vector(const VectorExpression<T2, A> &expr)
: isize(0) {
isize = expr.Size(0);
data_m.resize(isize);
for (int i = 0; i < isize; i++) {
data_m[i] = expr(i);
}
}
