> I think queue_t should be removed completely. We should just use
> odp_queue_t for internal interfaces as well. Looking at the code,
> using odp_queue_t would introduce 3 extra addition instructions and
> the impact to cache does not change. We can run l2fwd benchmark and
> compare the numbers. This would keep the code simple as well.
First, this comment highlights why one direct function call provides better
modularity than unnecessary exposition of an internal mid type. With
handle_to_qentry() we can change, or even remove, the mid type without need to
touch any of the call sites.
queue_entry_t *queue;
odp_buffer_hdr_t *buf_hdr;
- queue = qentry_from_int(queue_from_ext(handle));
+ queue = handle_to_qentry(handle);
Secondly, the internal queue pointer (or table index - format does not matter)
is needed to avoid doing the same conversion multiple times. We are inside
implementation, and it improves performance when the number of odp_queue_t to
pointer (qentry) conversions per packet is minimized.
After this basic clean up is merged, I'll look into the queue interface
definition and optimize it. This current version is just a dummy, copy-paste
replacement of qentry->xxx with fn_queue->xxx(). Which is OK for the first
stage, but certainly not optimal/final solution.
-Petri
