Performance degradation due to default initialisation of elements in standard containers

Question

(Yes, I know there is a question with almost the same title, but the answer was not satisfactory, see below)

EDIT Sorry, the original question didn\'t u

Answer 1

With g++ 4.5 I was able to realize an approximate 20% reduction in runtime from v0 (1.0s to 0.8s) and slightly less from 0.95s to 0.8s for v1 by using a generator to construct directly:

struct Generator : public std::iterator<std::forward_iterator_tag, int>
{
    explicit Generator(int start) : value_(start) { }
    void operator++() { ++value_; }
    int operator*() const { return value_; }

    bool operator!=(Generator other) const { return value_ != other.value_; }

    int value_;
};

int main()
{
    const int n = 100000000;
    std::vector<int> v(Generator(0), Generator(n));

    return 0;
}

Answer 2

I'm actually going to suggest in this case to roll your own container or look for alternatives since with the way I see it, your inherent problem is not with standard containers default constructing elements. It's with trying to use a variable-capacity container for one whose capacity can be determined upon construction.

There is no instance where the standard library needlessly default constructs elements. vector only does so for its fill constructor and resize, both of which are conceptually required for a general-purpose container since the point of those is to resize the container to contain valid elements. Meanwhile it's simple enough to do this:

T* mem = static_cast<T*>(malloc(num * sizeof(T)));
for (int j=0; j < num; ++j)
     new (mem + j) T(...); // meaningfully construct T
...
for (int j=0; j < num; ++j)
     mem[j].~T();         // destroy T
free(mem);

... and then build an exception-safe RAII-conforming container out of the code above. And that's what I suggest in your case since if default construction is wasteful enough to be non-negligible in a fill constructor context to the point where the alternative reserve and push_back or emplace_back is equally inadequate, then chances are that even a container treating its capacity and size as a variable is a non-negligible overhead, at which point you are more than justified to seek out something else, including rolling your own thing from the concept above.

The standard library is pretty damned efficient for what it does in ways where it's incredibly difficult to match in apples to apples comparisons, but in this case your needs call for oranges rather than apples. And in such cases, it often starts to become easier to just reach for an orange directly rather than trying to transform an apple to become an orange.

Answer 3

Okay, here is what I've learned since asking this question.

Q1 (Is there any legal way to use a standard library container which would give these latter timings?) Yes to some degree, as shown in the answers by Mark and Evgeny. The method of providing a generator to the constructor of std::vector elides the default construction.

Q2 (Is there any legal way to use a standard library container which would give these latter timings in multi-threaded situations?) No, I don't think so. The reason is that on construction any standard-compliant container must initialise its elements, already to ensure that the call to the element destructors (upon destruction or resizing of the container) is well-formed. As the std library containers do not support the usage of multi-threading in constructing their elements, the trick of Q1 cannot be replicated here, so we cannot elide the default construction.

Thus, if we want to use C++ for high-performance computing our options are somewhat limited when it comes to managing large amounts of data. We can

1 declare a container object and, in the same compilation unit, immediately fill it (concurrently), when the compiler hopefully optimizes the initialization at construction away;

2 resort to new[] and delete[] or even malloc() and free(), when all the memory management and, in the latter case, construction of elements is our responsibility and our potential usage of the C++ standard library very limited.

3 trick a std::vector to not initialise its elements using a custom unitialised_allocator that elides the default construction. Following the ideas of Jared Hoberock such an allocator could look like this (see also here):

// based on a design by Jared Hoberock
// edited (Walter) 10-May-2013, 23-Apr-2014
template<typename T, typename base_allocator = std::allocator<T> >
struct uninitialised_allocator
  : base_allocator
{
  static_assert(std::is_same<T,typename base_allocator::value_type>::value,
                "allocator::value_type mismatch");

  template<typename U>
  using base_t =
    typename std::allocator_traits<base_allocator>::template rebind_alloc<U>;

  // rebind to base_t<U> for all U!=T: we won't leave other types uninitialised!
  template<typename U>
  struct rebind
  {
    typedef typename
    std::conditional<std::is_same<T,U>::value,
                     uninitialised_allocator, base_t<U> >::type other; 
  }

  // elide trivial default construction of objects of type T only
  template<typename U>
  typename std::enable_if<std::is_same<T,U>::value && 
                          std::is_trivially_default_constructible<U>::value>::type
  construct(U*) {}

  // elide trivial default destruction of objects of type T only
  template<typename U>
  typename std::enable_if<std::is_same<T,U>::value && 
                          std::is_trivially_destructible<U>::value>::type
  destroy(U*) {}

  // forward everything else to the base
  using base_allocator::construct;
  using base_allocator::destroy;
};

Then an unitialised_vector<> template could be defined like this:

template<typename T, typename base_allocator = std::allocator<T>>
using uninitialised_vector = std::vector<T,uninitialised_allocator<T,base_allocator>>;

and we can still use almost all of the standard library's functionality. Though it must be said that the uninitialised_allocator, and hence by implication the unitialised_vector are not standard compliant, because its elements are not default constructed (e.g. a vector<int> will not have all 0 after construction).

When using this tool for my little test problem, I get excellent results:

timing vector::vector(n) + set_v0();
n=10000 time: 3.7e-05 sec
n=100000 time: 0.000334 sec
n=1000000 time: 0.002926 sec
n=10000000 time: 0.028649 sec
n=100000000 time: 0.293433 sec

timing vector::vector() + vector::reserve() + set_v1();
n=10000 time: 2e-05 sec
n=100000 time: 0.000178 sec
n=1000000 time: 0.001781 sec
n=10000000 time: 0.020922 sec
n=100000000 time: 0.428243 sec

timing vector::vector() + vector::reserve() + set_v0();
n=10000 time: 9e-06 sec
n=100000 time: 7.3e-05 sec
n=1000000 time: 0.000821 sec
n=10000000 time: 0.011685 sec
n=100000000 time: 0.291055 sec

timing vector::vector(n) + omp parllel set_v0();
n=10000 time: 0.00044 sec
n=100000 time: 0.000183 sec
n=1000000 time: 0.000793 sec
n=10000000 time: 0.00892 sec
n=100000000 time: 0.088051 sec

timing vector::vector() + vector::reserve() + omp parallel set_v0();
n=10000 time: 0.000192 sec
n=100000 time: 0.000202 sec
n=1000000 time: 0.00067 sec
n=10000000 time: 0.008596 sec
n=100000000 time: 0.088045 sec

when there is no difference any more between the cheating and "legal" versions.

Answer 4

boost::transformed

For single-thread version, you may use boost::transformed. It has:

Returned Range Category: The range category of rng.

Which mean, that if you would give Random Access Range to boost::transformed, it would return Random Access Range, what would allow vector's constructor to pre-allocate required amount of memory.

You may use it as follows:

const auto &gen = irange(0,1<<10) | transformed([](int x)
{
    return exp(Value{x});
});
vector<Value> v(begin(gen),end(gen));

LIVE DEMO

#define BOOST_RESULT_OF_USE_DECLTYPE 
#include <boost/range/adaptor/transformed.hpp>
#include <boost/container/vector.hpp>
#include <boost/range/irange.hpp>
#include <boost/progress.hpp>
#include <boost/range.hpp>
#include <iterator>
#include <iostream>
#include <ostream>
#include <string>
#include <vector>
#include <array>


using namespace std;
using namespace boost;
using namespace adaptors;

#define let const auto&

template<typename T>
void dazzle_optimizer(T &t)
{
    auto volatile dummy = &t; (void)dummy;
}

// _______________________________________ //

using Value = array<int,1 << 16>;
using Vector = container::vector<Value>;

let transformer = [](int x)
{
    return Value{{x}};
};
let indicies = irange(0,1<<10);

// _______________________________________ //

void random_access()
{
    let gen = indicies | transformed(transformer);
    Vector v(boost::begin(gen), boost::end(gen));
    dazzle_optimizer(v);
}

template<bool reserve>
void single_pass()
{
    Vector v;
    if(reserve)
        v.reserve(size(indicies));
    for(let i : indicies)
        v.push_back(transformer(i));
    dazzle_optimizer(v);
}

void cheating()
{
    Vector v;
    v.reserve(size(indicies));
    for(let i : indicies)
        v[i]=transformer(i);
    dazzle_optimizer(v);
}

// _______________________________________ //

int main()
{
    struct
    {
        const char *name;
        void (*fun)();
    } const tests [] =
    {
        {"single_pass, no reserve",&single_pass<false>},
        {"single_pass, reserve",&single_pass<true>},
        {"cheating reserve",&cheating},
        {"random_access",&random_access}
    };
    for(let i : irange(0,3))
        for(let test : tests)
            progress_timer(), // LWS does not support auto_cpu_timer
                (void)i,
                test.fun(),
                cout << test.name << endl;

}