Why TVMArrayAlloc API don't support strides?

Questions
1

Two Questions:

  1. TVMArrayAlloc API don’t support alloc DLTensor with strides, although we can set strides outside, but the data memory can’t change again if strides is larger than shape.

  2. If we export an op library and using it in c/c++ runtime, and bind input/output with decl_buffer that using strides option, should we care about strides in python compute primitives?

Thanks.