Wrong sign in a single-gate-circuit statevector? about cuquantum

[yapolyak]

Oh I've just managed to reproduce the result the State.compute() is giving me! If I transpose the unitary, I get the same result! But why is this the case, and should I transpose every unitary that goes into State?

[yapolyak]

I now recall that you do transpose Pauli unitaries in your example, and have been ignoring that... This probably is due to necessity of using 'F'-style ordering, right?

[yapolyak]

Another update - if I try and provide gate_ry strides explicitly (by using cupy.ndarray.strides() (accounted for the data type byte length) to cutn.state_apply_tensor(), it gives me a wrong result anyway, unless I transpose the array.

However, if I use order='C', don't transpose and Don't provide strides (pass 0) I get the right result. I think I understand why this would work (order='C' is doing the same thing as transposing here), but why if I provide explicit strides it fails? Specifically, I calculate the strides as follows:

gate_ry = cp.asarray([[ 0.33873792+0.j, -0.94088077+0.j],
                      [ 0.94088077+0.j,  0.33873792+0.j]]).astype(dtype='complex128', order='C')
gate_ry_strides = tuple(int(stride/16) for stride in gate_ry.strides)