On Fri, Aug 7, 2009 at 15:43, Jeff Squyres<jsquyres_at_[hidden]> wrote:
> On Jul 29, 2009, at 4:53 PM, Bert Wesarg wrote:
>
>> I found my access code and would commit patches 1-4 in the near
>> future. For patch 5 I would like to get an ACK, because I'm not sure
>> if this is needed.
>>
>
>
> For patch 5, is your concern looping over calling plpa_get_socket_id() from
> 0 to the max potentially far too wasteful as opposed to looping from min to
> max?
get_socket_id() is the best way to loop over all known sockets,
because it maps from ith socjet to socket id, and therefore avoid a
loop over all possible socket ids, i.e. 0..socket_id_max.
>
> It does strike me that we should have provided the minimum from the
> beginning for exactly this reason... Â Additionally, perhaps we should have
> used a dynamically-sized bitmask instead of a fixed-size one; then we could
> have returned all valid socket ID's in a single call (with a loop of macro
> calls to check bits to see which exact IDs were valid). Â However, this is
> likely a conversation to take forward in the merging of PLPA and libtopology
> (no one objected to the idea here on the list, so I took the liberty of
> having some more off-list conversations with the libtopology authors).
I will comment on this later.
Bert
>
> --
> Jeff Squyres
> jsquyres_at_[hidden]
>
> _______________________________________________
> plpa-users mailing list
> plpa-users_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/plpa-users
>
|
![[Flag]](../../../../../community/mirrors/flag-us.gif){kind=small}