using std::valarray and std::slice to copy stridden data

Question

I'm trying efficiently to copy stridden data. Consider following function signature

void foo(Array& input, Buffer& U, Buffer& V)

where Buffer and Array are

using Buffer = std::vector<uint8_t>;
using Array = std::valarray<uint8_t>;

Now, depending on constness of input I can slice the valarray and get another valarray or slice_array. Getting sliced valarray harms performance (copy). Is there a way to copy sliced data using the slice_array?

My naïve implementation

void foo(Array& input, Buffer& U, Buffer& V)
{
    Array Uslice = input[std::slice(0, input.size() / 2, 2)];
    Array Vslice = input[std::slice(0 + 1, input.size() / 2, 2)];

    std::copy(std::begin(Uslice), std::end(Uslice), U.begin());
    std::copy(std::begin(Vslice), std::end(Vslice), V.begin());
}

However, this implementation is slower than just looping over the input and assigning values by index.

BTW, I'm not bound to valarray, in case you have better ideas what container to use

EDIT001: Second thought, Range-V3 may be handy in this case, with its ranges::v3::view::stride, however I cant find any example of how to use it

EDIT002: Managed to get ranges working. It is terribly slow, 20 times slower than simple for loop.

void foo(Buffer& input, Buffer& U, Buffer& V)
{
    V = ranges::move(input | ranges::v3::view::slice(0ull, input.size()) | ranges::v3::view::stride(2));
    U = ranges::move(input | ranges::v3::view::slice(1ull, input.size()) | ranges::v3::view::stride(2));
}

What I'm doing wrong?