[Nek5000-users] New Recycling BC based on interpolation

[nek5000-users at lists.mcs.anl.gov]

Hi Guys,

I worked on the interpolation wrappers over the last months and I am not sure if the new recycling BC is compatible to the latest repo-version?

Paul/Aleks: can you comment on that? 


Now, in the case of CHT (nelgt>nelgv) findpts_eval() won't work! I guess the size of the input field has to match the dimensions of XM1,YM1,ZM1 specified in inpts_setup(). 

I will talk to James if we can come up with an easy fix for that.


Stefan  



-----Original Message-----
From: nek5000-users-bounces at lists.mcs.anl.gov on behalf of nek5000-users at lists.mcs.anl.gov
Sent: Sun 7/18/2010 13:53
To: nek5000-users at lists.mcs.anl.gov
Subject: Re: [Nek5000-users] New Recycling BC based on interpolation
 


Hi Markus,

this stuff is all brand new and you are in the leading edge
of the testing (since none of us are exercising it in our
own applications), so there will be a bit of development as
we go forward.  We can readily fix interp_v (but I advise
using it now in order to identify other issues).

interpv is _much_ more expensive than gs_op.

However, if one takes 1 part in a million and the other 1
part in 10,000 of the run time, then both are negligible.
So the question is really relative to the wall-clock time.

Paul



On Sun, 18 Jul 2010, nek5000-users at lists.mcs.anl.gov wrote:

> Hi,
>
> no, I haven't done any speed comparison. So I take it there is no principal 
> difference between intpts and gs_op in terms of computational cost?
>
> I was referring to line 172f of the new recycling BC code:
> "
>      if (nelgt.ne.nelgv) call exitti
>     $   ('ABORT: interp_v() nelgt.ne.nelgv not yet supported!$',nelgv)
> "
> and was wondering if that is due to something in intpts and if so, if the 
> same constraint would apply to findpts.
>
> Markus
>
>
> nek5000-users at lists.mcs.anl.gov wrote:
>> 
>> Hi Markus,
>> 
>> I don't know why this wouldn't work for nelgt > nelgv.
>> 
>> I would say we don't know the performance hit yet - I think
>> someone is looking into this question.  Performance will of course vary 
>> significantly from problem to problem and platform
>> to platform.   I suspect that it should not be a major hit ---
>> have you observed otherwise?
>> 
>> Paul
>> 
>> 
>> On Sat, 17 Jul 2010, nek5000-users at lists.mcs.anl.gov wrote:
>> 
>>> Hi Aleks,
>>> 
>>> thanks for the code, I looked over it and have two questions:
>>> -at some point, it is mentioned that this method does not work if 
>>> nelgt>nelgv. Why is that? I am asking because this will be applied to a 
>>> conjugate HT problem (although the recycling part is purely concerning 
>>> flow+convection). Will this problem also occur when only findpts is used 
>>> instead of intpts?
>>> -is there a performance difference between calling intpts as opposed to 
>>> gs_op (as is done in the turbJet example) every time step for large scale 
>>> production runs? The latter is what I used in my version.
>>> 
>>> Thanks,
>>> Markus
>>> 
>>> 
>>> nek5000-users at lists.mcs.anl.gov wrote:
>>>> Hi Michael and Markus,
>>>> 
>>>> Here is a test case for an unsteady pipe flow with Michael's 
>>>> GambitPipe.rea file which sets inlet Recycling BCs based on interpolation 
>>>> from a cross-section downstream that you specify by a multiple of the 
>>>> inlet normal:
>>>> 
>>>> http://www.mcs.anl.gov/~obabko/z4.tgz
>>>> 
>>>> 
>>>> Let me know if you have any questions.
>>>> Best,
>>>> Aleks
>>>> 
>>>> 
>>>> 
>>>> 
>>>> ----- Original Message -----
>>>> From: nek5000-users at lists.mcs.anl.gov
>>>> To: nek5000-users at lists.mcs.anl.gov
>>>> Sent: Thursday, July 8, 2010 5:06:48 PM GMT -06:00 US/Canada Central
>>>> Subject: Re: [Nek5000-users] New Recycling BC method?
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> No worries, just checkin in to see how it was going, thanks for the 
>>>> update! Look forward to trying it out!
>>>> 
>>>> - Michael _______________________________________________
>>>> Nek5000-users mailing list
>>>> Nek5000-users at lists.mcs.anl.gov
>>>> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
>>>> 
>>> _______________________________________________
>>> Nek5000-users mailing list
>>> Nek5000-users at lists.mcs.anl.gov
>>> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
>>> 
>> _______________________________________________
>> Nek5000-users mailing list
>> Nek5000-users at lists.mcs.anl.gov
>> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
>> 
> _______________________________________________
> Nek5000-users mailing list
> Nek5000-users at lists.mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
>
_______________________________________________
Nek5000-users mailing list
Nek5000-users at lists.mcs.anl.gov
https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users