Roland, Yes, that all makes sense. Will change my code appropriately. Thanks! Hoang-Nam Nguyen
Roland Dreier
<[EMAIL PROTECTED]
m> To
[EMAIL PROTECTED]
07.03.2006 22:39 cc
[email protected],
Christoph Raisch/Germany/[EMAIL
PROTECTED],
Hoang-Nam Nguyen/Germany/[EMAIL
PROTECTED]
Subject
Re: [openib-general] [PATCH 11/17]
ehca: completion queue
> + list_add(&qp->list_entries, &cq->qp_hashtab[key]);
You could save some memory by making qp_hashtab a struct hlist instead
of a struct list_head.
> + unsigned int key = real_qp_num%QP_HASHTAB_LEN;
This seems like a slightly funny hash function: you're doing an
integer divide in the poll_cq fast path, since QP_HASHTAB_LEN is 7.
How would making QP_HASHTAB_LEN be 8 affect the distribution (that way
the % could become a bitwise &)?
- R.
_______________________________________________
openib-general mailing list
[email protected]
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
