Random dependent variables using QtConcurrent - c++

Question

I ran into a problem of design. Prepare, as there will be a lot to come, and perhaps it is a too complex problem for SO.
For people familiar with R this might help:
I created a package in R for creating correlations between variables (for the purpose of simulations). I created a multiple threads version to allow parallelism. While R makes it seem embarrassingly parallel, I almost reconsider this statement in C++.
Sample a 100000 by 2 random data matrix.
Split it into a list of (100,000 / 200) = 500 splits of 200 rows by 2 columns data matrices.
Using a multi-core list apply function (mclapply, which can handle matrices as "item" in a list), I was able to send these smaller data matrices with 2 dimensions to the same function, though in different threads.
The function randomly rearranged the first column conditionally on the second column to produce a certain correlation. It then returned the rearranged matrix. I am not asking for this part, I am just mentioning this to outrule solutions that skip steps by independently creating x and y.
The list apply function automatically glues together these 500 splits of 200 rows by 2 columns into one big matrix (100,000 by 2).
listdata:
[[1]]
[1] [2]
... ...
... ...
[[2]]
[1] [2]
... ...
... ...
Applying:
mclapply(listdata, myFunction, cores=4)
Result:
[1] [2]
... ...
... ...
... ...
... ...
Randomness
I want to have a thread-safe, statistically random (results unrelated to the pseudo randomness) function yielding aggregated dependent x and y vectors.
Randomness is difficult since there usually is a global state for a random generator, which is not possible to share in a parallel situation.
Some of my solutions showed that the threads all used similar values, which does not happen using R.
Qt
I am using Qt, and I am trying to use QtConcurrent for creating the right (convenient for user) amount of threads, given a variable amount of splits.
I'll first say that I did not find a way to be able to pass multiple arguments to a function in a QtConcurrent. They all seem to be focused on mapping, without allowing extra arguments.
One attempt was the following:
QList< QVector<double> > list_of_one_split;
QList< QList< QVector<double> > list_containing_all_splits;
QVector<double> vector_of_doubles(10);
qFill(vector_of_doubles.begin(), vector_of_doubles.end(), 1); // not random
list_of_a_split.append(vector_of_doubles); // x variable split
list_of_a_split.append(vector_of_doubles); // y variable split
list_containing_all_splits.append(list_of_a_split) // first split
list_containing_all_splits.append(list_of_a_split) // second split
QtFuture<void> res = `QtConcurrent::map(list_containing_all_splits.begin(),
list_containing_all_splits.end(),
mapFunction)`
QList< QVector<double> > mapFunction(QList< QVector<double> > &list_of_a_split)
{
// random arranging one column on other
return list_of_a_split;
}
For some reason this return did not seem to affect the original list.
Another way I tried was to generate the y within the mapper function, giving it X as input, and returning y into a new variable using QtConcurrent:mappedReduced(), mapping over x in split form, producing y, return y split, and using reduce to merge it together into QVector<double> y. This has serious random issues, as I am not able to send a generator as extra argument.
I am just showing my efforts here, but feel free to approach the problem from any way (though same results) you like.
The goal in the end:
// Whether or not packed in a QList< >:
QVector<double> x = {..., ..., ..., ...};
QVector<double> y = {..., ..., ..., ...}; // dependent on x