Difference bwtween v4l2 userptr and r\w mode

I am trying to use v4l2 with shared memory.

userptr mode uses queue , where queue us limited to hold maximum 32 buffers.

to overcome the limitation,i thought about using r\w mode.
the idea is to manage a private queue with 300 pointers to buffers.
every time i call v4l2 \ read, i will send it a pointer to the next buffer

question is: why is userptr considered less managable than userptr ?
(documantioation mentioned time stamps, but i i not sure what it means)

hello regine.issan,

it’s depends-on V4L2 capturing modes and CUDA memory management.

may I also have more details for such approach?
for example, could you please describe what’t the real use-case.