Well... Considering the figures it is clear that either of your patches can
make a big difference to Kannel's performance.
MHO is that we go for the least complex patch, if test results appear more
or less the same between the two.

== Rene


-----Original Message-----
From: [email protected] [mailto:[email protected]] On
Behalf Of Alexander Malysh
Sent: Tuesday, 07 August, 2012 15:12
To: Stipe Tolj
Cc: [email protected]
Subject: Re: [PATCH] Lightning speed Dict hash table

Hi Stipe,

this is really complex patch for the hash issue, but it could be made easier
and almost the same speed.
I'm not really convenient that we need to apply this complex patch...

Here are my simple tests and patch attached:

        Command used:

time ./test/test_dict 

-- old
2012-08-07 14:47:49 [57291] [0] DEBUG: Dict populate phase.
2012-08-07 14:48:16 [57291] [0] INFO: ok, got 200000 entries in dict1.
2012-08-07 14:48:16 [57291] [0] INFO: ok, got 200000 entries in dict2.
2012-08-07 14:48:16 [57291] [0] DEBUG: Dict lookup phase.

real    0m41.842s
user    0m40.680s
sys     0m1.078s

-- sedgewick (my patch)
2012-08-07 14:47:01 [57016] [0] DEBUG: Dict populate phase.
2012-08-07 14:47:03 [57016] [0] INFO: ok, got 200000 entries in dict1.
2012-08-07 14:47:03 [57016] [0] INFO: ok, got 200000 entries in dict2.
2012-08-07 14:47:03 [57016] [0] DEBUG: Dict lookup phase.

real    0m4.160s
user    0m2.998s
sys     0m0.977s

real    0m3.797s
user    0m2.738s
sys     0m1.015s

real    0m3.732s
user    0m2.767s
sys     0m0.938s

-- Bob Jenkins
2012-08-07 14:50:32 [58470] [0] DEBUG: Dict populate phase.
2012-08-07 14:50:34 [58470] [0] INFO: ok, got 200000 entries in dict1.
2012-08-07 14:50:34 [58470] [0] INFO: ok, got 200000 entries in dict2.
2012-08-07 14:50:34 [58470] [0] DEBUG: Dict lookup phase.

real    0m3.747s
user    0m2.682s
sys     0m0.919s

real    0m3.661s
user    0m2.692s
sys     0m0.946s

real    0m3.522s
user    0m2.606s
sys     0m0.908s

Alex


Reply via email to