[Net-snmp-patches] [ net-snmp-Patches-3195537 ] binary_array container sort optimization

[SourceForge.net <no...@so...>]

Patches item #3195537, was opened at 2011-02-28 22:40
Message generated for change (Comment added) made by bvassche
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=312694&aid=3195537&group_id=12694

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: None
Group: None
>Status: Closed
>Resolution: Accepted
Priority: 5
Private: No
Submitted By: Stephen Hemminger ()
Assigned to: Nobody/Anonymous (nobody)
Summary: binary_array container sort optimization

Initial Comment:
Put the simple comparison of index before the more
costly comparison function when evaluating conditionals

----------------------------------------------------------------------

>Comment By:  (bvassche)
Date: 2011-03-07 18:20

Message:
Applied as r20077 on the trunk and the 5.6 branch - thanks for the patch.

----------------------------------------------------------------------

Comment By:  (bvassche)
Date: 2011-03-02 08:10

Message:
What I would prefer (if this makes sense) is to replace the array used by
the container_binary_array.c implementation by a red-black tree. Inserting
in and removing from a red-black tree is much more scalable than resorting
an array upon every insertion and removal. Is there a BSD-licensed
red-black tree implementation available somewhere ?

----------------------------------------------------------------------

Comment By:  (bvassche)
Date: 2011-03-01 21:04

Message:
I would also prefer to see the qsort() function provided by the C library
to be invoked here. Unfortunately the function type
netsnmp_container_compare is not what qsort() expects: functions of type
netsnmp_container_compare are passed an element of array "data" while
qsort() passes pointers to elements of array "data" to its comparison
function.

----------------------------------------------------------------------

Comment By: Magnus Fromreide (magfr)
Date: 2011-03-01 00:40

Message:
I think this is a false optimization.
i is the lower limit so it will only be true once we are down to a single
element range and that should preferably not happen.
Now, the real question is probably why we are using a homebrew qsort
instead of the standard library implementation.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=312694&aid=3195537&group_id=12694