For trivial types, the QVector::QVector(int N) constructor gives a vector filled with N default-constructed values. E.g., QVector<double> vec(100) gives a 100-element vector with each element pre-initialized to 0.0. I suggest adding an additional constructor of the form QVector::QVector(int N,Qt::Initialization) that would allow the vector to contain N uninitialized values (trivial types only).

As an example use case, consider a vector to contain 10M doubles. Currently, there are two options:

int N = 10000000;

//slow
QVector<double> vec;
vec.reserve(N);
for(int i=0; i<N; i++)
     vec.append(...);

//faster
QVector<double> vec(N); //all N elements initialized to 0.0
for(int i=0; i<N; i++)
     vec[i] = ...;

The first case is slow; the overhead of append() is significant. The second case is reasonably fast, but there is still the overhead of initializing all N values to 0.0, even though they will be immediately overwritten in the next loop. When N is large (and particularly if this code is in a loop), the overhead becomes significant.

Instead, for trivial types, an option like what follows would eliminate that extra overhead for cases in which the container will be manually initialized:

//fastest
QVector<double> vec(N,Qt::Uninitialized); //all N elements uninitialized
for(int i=0; i<N; i++)
     vec[i] = ...;

QVector is an example, but this could apply to other container classes as well.