[MITgcm-support] Optimization
Matthew Mazloff
mmazloff at ucsd.edu
Fri Jul 9 14:49:07 EDT 2010
Hi David,
I believe config_summary is only called once near the end of the
initialization.
actually many of the routines you show are initialization routines.
I am not sure about your profile below, but I have a hunch you are not
profiling a long enough run. Profile a full run -- or at least an
hour -- if you profile a ten minute run all you will see is that it
takes a few minutes to initialize the model.
-Matt
On Jul 9, 2010, at 11:36 AM, David Lewis wrote:
> Hello all,
>
> I'm running a global circulation routine on mitgcm on 32 processor
> and am looking to see if I can speed up the runtime. I had an IT
> person help and she sent me part of the profile (below). Most of
> the time seems to go into the routine "config_summary", which
> appears to just be IO. Is there any way I can reduce the number of
> times this routine is called? Does this have anything to do with
> the generation of STDOUT files?
>
> Thanks for your help.
>
> David
>
> =========================================================
> % cumulative self self total
> time seconds seconds calls ms/call ms/call name
> 11.3 65.49 65.49 3103165 0.02
> 0.02 .config_summary [3]
> 6.5 103.37
> 37.88 ._tag_waiting [6]
> 4.8 131.09
> 27.72 ._barrier_onnode [10]
> 3.9 153.69 22.60 2945398 0.01
> 0.01 .initialise_varia [289]
> 3.2 172.27
> 18.58 .gad_fluxlimit_adv_x [4]
> 3.1 190.12 17.85 1104516 0.02 0.03 .gad_advection
> [8]
> 3.0 207.63 17.51 3155760 0.01
> 0.01 .ini_cartesian_grid [26]
> 2.8 223.97
> 16.34 .gad_del2 [11]
> 2.8 240.18 16.21 3155760 0.01
> 0.01 .poe_remote_main [15]
> 2.8 256.35 16.17 52596 0.31 1.55 .ini_cg2d [2]
> 2.7 271.76 15.41 2103840 0.01 0.01 .find_rho_2d
> [12]
> 2.6 286.68 14.92 2156436 0.01
> 0.02 .calc_oce_mxlayer [7]
> 2.2 299.42
> 12.74 .ini_depths [13]
> 2.2 311.92
> 12.50 .__mcount [14]
>
> Hotmail has tools for the New Busy. Search, chat and e-mail from
> your inbox. Learn more. <ATT00001.txt>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20100709/b297fe9a/attachment.htm>
More information about the MITgcm-support
mailing list