Bug 88 – Ipv4Address operator ==, !=, < should be inline for performance

Bug 88 - Ipv4Address operator ==, !=, < should be inline for performance


Summary:	Ipv4Address operator ==, !=, < should be inline for performance

Status:	RESOLVED FIXED

Product:	ns-3
Classification:	Unclassified
Component:	network
Version:	pre-release
Hardware:	All All

Importance:	P3 enhancement
Assigned To:	ns-bugs

URL:
Keywords:

Depends on:
Blocks:
	Show dependency tree / graph

Reported:	2007-10-10 13:07 UTC by Gustavo J. A. M. Carneiro
Modified:	2007-10-11 10:40 UTC (History)
CC List:	0 users

See Also:

Attachments
make critical Ipv4Address methods/operators inline (2.67 KB, patch) 2007-10-10 13:40 UTC, Gustavo J. A. M. Carneiro	Details \| Diff
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.

Description Gustavo J. A. M. Carneiro 2007-10-10 13:07:21 UTC

As seen in http://www.nsnam.org/bugzilla/attachment.cgi?id=76, Ipv4Address operators are consuming significant time in my OLSR simulations.

Making them inline improves a script of mine simulation time from 3 to 2.5 seconds.  I would attach a patch, but bugzilla is temporarily broken and won't let me.

Comment 1 Gustavo J. A. M. Carneiro 2007-10-10 13:19:38 UTC

hm.. I made a mistake in the previous measurements.  Correct tests follow:


gjc@spectrum:ns-3-olsr-swlan$ time ./build/optimized/utils/bench-olsr  --num-nodes 15
0:38.7298:0:38.7298
Average Delay: 2.93073 s
Packet Loss: 64.5603 %

real    0m11.236s
user    0m10.857s
sys     0m0.100s
gjc@spectrum:ns-3-olsr-swlan$ time ./build/optimized/utils/bench-olsr  --num-nodes 15
0:38.7298:0:38.7298
Average Delay: 2.82314 s
Packet Loss: 63.3327 %

real    0m15.627s
user    0m15.253s
sys     0m0.084s
gjc@spectrum:ns-3-olsr-swlan$ time ./build/optimized/utils/bench-olsr  --num-nodes 15
0:38.7298:0:38.7298
Average Delay: 2.92139 s
Packet Loss: 64.3521 %

real    0m12.926s
user    0m12.233s
sys     0m0.040s
gjc@spectrum:ns-3-olsr-swlan$ hg revert src/node/ipv4-address.{h,cc} 
gjc@spectrum:ns-3-olsr-swlan$ 
gjc@spectrum:ns-3-olsr-swlan$ 
gjc@spectrum:ns-3-olsr-swlan$ 
gjc@spectrum:ns-3-olsr-swlan$ 
gjc@spectrum:ns-3-olsr-swlan$ 
gjc@spectrum:ns-3-olsr-swlan$ 
gjc@spectrum:ns-3-olsr-swlan$ 
gjc@spectrum:ns-3-olsr-swlan$ 
gjc@spectrum:ns-3-olsr-swlan$ 
gjc@spectrum:ns-3-olsr-swlan$ time ./build/optimized/utils/bench-olsr  --num-nodes 15
0:38.7298:0:38.7298
Average Delay: 2.8314 s
Packet Loss: 63.4598 %

real    0m17.499s
user    0m16.717s
sys     0m0.096s
gjc@spectrum:ns-3-olsr-swlan$ time ./build/optimized/utils/bench-olsr  --num-nodes 15
0:38.7298:0:38.7298
Average Delay: 2.82491 s
Packet Loss: 63.0099 %

real    0m15.068s
user    0m14.749s
sys     0m0.068s
gjc@spectrum:ns-3-olsr-swlan$ time ./build/optimized/utils/bench-olsr  --num-nodes 15
0:38.7298:0:38.7298
Average Delay: 2.91977 s
Packet Loss: 64.2376 %

real    0m15.457s
user    0m14.805s
sys     0m0.096s


Average user time with patch (inline operators): 12.781 s
Average user time without patch: 15.424 s (+20%)

Comment 2 Gustavo J. A. M. Carneiro 2007-10-10 13:40:43 UTC

Created attachment 78 [details]
make critical Ipv4Address methods/operators inline

Comment 3 Mathieu Lacage 2007-10-11 00:23:37 UTC

ok for me.