Reading data from neuroglancer in the correct order about tensorstore HOT 3 CLOSED

OK. Thank you very much for this information.

Is there a difference in efficiency between my variant and using Transpose()?

No.

from tensorstore.

There is both the logical dimension order and the physical memory layout of an in-memory array. These are independent in tensorstore.

It sounds like you want x to be contiguous in memory, i.e. xyzc Fortran/column-major order, or equivalently, czyx c-order/row-major.

The neuroglancer_precomputed driver has the dimension order x, y, z, channel by default.

Therefore you can either read it directly to a Fortran order array (as in your snippet) or transpose it, e.g. with Dims("channel","z","y","x").Transpose() and then read it to a c-order array.

from tensorstore.

OK. Thank you very much for this information.

Is there a difference in efficiency between my variant and using Transpose()?

from tensorstore.