Need help with TV layout

[dukallis]

Can anybody help with the problem to pick correct value for parameters? I have the following code snippet:

#include "cute/layout.hpp"
#include "cute/tensor.hpp"

using namespace cute;

int main() {
    // (64, (4, 2)) : (1, (128, 64))
    #if defined(MANUAL)
    auto tv_layout = Layout<Shape<Shape<_64>, Shape<_4, _2>>,
                           Stride<Stride<_1>, Stride<_128, _64>>>{};

    auto tiler = Shape<_128, _4>{};

    TiledCopy copy_b = TiledCopy<Copy_Atom<UniversalCopy<float>, float>,
                                 decltype(tv_layout), decltype(tiler)>{};
    #else
    auto copy_b = make_tiled_copy(Copy_Atom<UniversalCopy<float>, float>{},
                                     Layout<Shape<_64>>{},
                                     Layout<Shape<_1, _4>>{}); // Closest I could get. 
                                                               // I just can't understand how to repeat the pattern vertically once more 
    #endif

    print_latex(copy_b);
    print(copy_b);
}

I need copy_b from both conditional compilation branches to produce the same picture obtained from output of print_latex. With the values above I have one "brick" of that pattern and I just need the next one. Can't understand how to get there.

[dukallis]

I believe that's impossible