How is thread data mapping in cute mma atom TN layout?

[WhoisZihan]

I've read the mma_atom document, and I am confused with the TN layout in the figure below.

Specifically the A layout in TN is

// (T8,V4) -> (m,k)
  using ALayout = Layout<Shape <_8,_4>,
                         Stride<_1,_8>>

But A is transposed, and is K-major. Why the stride is <_1, _8> instead of <_4, _1>?

Suppose I want to fetch (T1, V2), the calculated coordinate is 1x1+2x8=17, but the memory index is actually 4+2=6, how are they linked together?