[MITgcm-support] The way to accelerate the calculation speed or to expand deltaT of MITgcm

[姊佹湅]

Dear sir:
       Thank you for your advice to my problem.But in my experiment gravity waves and non-hydrostatic effects are important.So I can't shut it down.Is there anything else I can do to expand the deltaT?And now I met a new problem.I set the deltaT=60s,and I have calculated about 30000 steps in my test,there is nothing wrong.So I started the calculating officailly,But after calculating about 140000 steps,the calculation was stopped,And the false information outputed is as follow:
 
p34_17723:  p4_error: interrupt SIGFPE: 8
rm_l_34_17736: (109867.313526) net_send: could not write to fd=5, errno = 32
p35_17738:  p4_error: interrupt SIGFPE: 8
rm_l_35_17751: (109867.315028) net_send: could not write to fd=5, errno = 32
p37_17768:  p4_error: interrupt SIGFPE: 8
rm_l_37_17781: (109867.239080) net_send: could not write to fd=5, errno = 32
p39_17798:  p4_error: interrupt SIGFPE: 8
rm_l_39_17811: (109867.163699) net_send: could not write to fd=5, errno = 32
p41_17828:  p4_error: interrupt SIGFPE: 8
rm_l_41_17841: (109867.083061) net_send: could not write to fd=5, errno = 32
p15_17934:  p4_error: interrupt SIGx: 13
p20_9939:  p4_error: interrupt SIGx: 13
p17_9876:  p4_error: net_recv read:  probable EOF on socket: 1
rm_l_17_9895: (109872.631953) net_send: could not write to fd=5, errno = 32
p19_9918:  p4_error: interrupt SIGx: 13
p33_17708:  p4_error: interrupt SIGFPE: 8
rm_l_33_17721: (109867.358075) net_send: could not write to fd=5, errno = 32
p27_14238:  p4_error: interrupt SIGFPE: 8
rm_l_27_14257: (109872.866971) net_send: could not write to fd=5, errno = 32
p28_14259:  p4_error: interrupt SIGFPE: 8
rm_l_28_14278: (109872.817565) net_send: could not write to fd=5, errno = 32
p40_17813:  p4_error: interrupt SIGFPE: 8
rm_l_40_17826: (109867.125186) net_send: could not write to fd=5, errno = 32
p38_17783:  p4_error: interrupt SIGFPE: 8
rm_l_38_17796: (109867.202276) net_send: could not write to fd=5, errno = 32
p32_17693:  p4_error: interrupt SIGFPE: 8
rm_l_32_17706: (109867.436921) net_send: could not write to fd=5, errno = 32
p29_14280:  p4_error: interrupt SIGFPE: 8
rm_l_29_14299: (109872.810261) net_send: could not write to fd=5, errno = 32
p3_12102:  p4_error: interrupt SIGx: 13
p31_14322:  p4_error: interrupt SIGFPE: 8
rm_l_31_14341: (109872.711552) net_send: could not write to fd=5, errno = 32
p30_14301:  p4_error: interrupt SIGx: 13
rm_l_30_14320: (109872.905991) net_send: could not write to fd=5, errno = 32  
 
So can you tell me what's the problem?Is the deltaT to long or something else?
Thank You
With Respects
                                                                                           Liang
--- 10年10月29日，周五, Jason Goodman <goodman_jason at wheatonma.edu> 写道：


发件人: Jason Goodman <goodman_jason at wheatonma.edu>
主题: Re: [MITgcm-support] The way to accelerate the calculation speed or to expand deltaT of MITgcm
收件人: mitgcm-support at mitgcm.org
日期: 2010年10月29日,周五,上午2:38




On Oct 28, 2010, at 5:27 AM, 姊佹湅 wrote:
> 
> > Dear Sir:
> >         I'm trying to use the MITgcm to run an experiment.To
> > simulate this experiment 
> > successfully it is necessary to set the grid interval to be very
> > small.Now the number of the horizontal grids is 114*104,and the
> > number of the vertical grids is 139.The minimum interval of the
> > horizontal grids is 5 km ,and the minimum interval of the vertical
> > grids is 10m.And now I find that the maximum deltaT in the data file
> > to keep the model steady is 60s. The minimum time that is needed is
> > 30 years,so the minimum number of timesteps (nTimeSteps in the data
> > file)is 16000000.Now I have used 48 cpu to run the model,and it
> > needs 210 days to finish the calculation.The time is too long!! So
> > can you tell me how to  set the model parameters or modify the model
> > code to  accelerate the calculation speed or to expand deltaT.

If your fluid is strongly stratified, the deltaT is probably limited by
the frequency of oscillation of internal gravity waves.  If you're not
interested in gravity wave dynamics or strong vertical motion, you can
experiment with setting

nonHydrostatic=.FALSE.,
implicitIntGravWave=.TRUE.,

I have not used the implicit gravity wave solver: it may have risks I
don't understand.  But if it works it may allow you to increase the
timestep toward the Courant limit for horizontal motions -- dt < dx/U =
2000 seconds or so.  (Viscosity may need to be adjusted to allow this.)

BUT, if you really do need to look at flow with strong vertical motions
where gravity waves or non-hydrostatic effects are important, I don't
think you can do much better than what you have now.  You will have to
either decrease the vertical resolution or be satisfied with a much
shorter simulation than 30 years.



      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20101029/69a47b6f/attachment-0001.htm>