[Developers] Recover SaveAndRestore variables with PUGH

[Erik Schnetter]

On Nov 2, 2006, at 18:11:11, Luca Baiotti wrote:

> Thomas Radke wrote:
>> Ian Hawke wrote:
>>
>>> On Wed, 2006-10-18 at 19:43 -0500, Erik Schnetter wrote:
>>>
>>>
>>>> On Oct 18, 2006, at 03:02:02, Luca Baiotti wrote:
>>>
>>>
>>>> I think the root of the problem is that evolution_method="none"  
>>>> does
>>>> not do anything to time levels.  It should instead copy the  
>>>> previous
>>>> time level into the current time level in the prestep bin.  The
>>>> current mechanism was implemented when people would use only a  
>>>> single
>>>> time level with stationary spacetimes; only with Carpet do people
>>>> want multiple time levels in that case, which breaks things.
>>>>
>>>> Ian, do you think that this would also solve the problem?
>>>
>>>
>>> It would work, and is rather more satisfactory than modifying MoL  
>>> (as
>>> SaveAndRestore variables just mean "these might be evolved by  
>>> some other
>>> thorn").
>>>
>>> I still believe that the only guaranteed way of ensuring consistency
>>> after recovery is to checkpoint all timelevels, even with PUGH,  
>>> because
>>> you never know what somebody's going to try and do.
>>
>>
>> If you don't want to solve the problem by modifying PUGH or MoL  
>> it's a
>> small thing for me to implement a new boolean parameter
>> IO::checkpoint_all_timelevels (defaulting to "no"?).
>>
>> Just let me know.
>
> This discussion has been quiet for a while; so I would assume that the
> last suggestion is supported by everyone.
>
> If this is the case, Thomas please implement your suggestion (it is  
> not
> so urgent, anyway).

I don't really; I think making the evolution method "none" work  
correctly (by not assuming that the new time level is taken from the  
pre-previous one) is superior.

I have placed a correspondingly modified copy of ADMBase onto Peyote  
in ~eschnett/ADMBase.tar.gz.  Couldyou test it?  It has some other,  
unrelated changes as well; if it solves your problem, we can clean it  
up.

-erik

-- 
Erik Schnetter <schnetter at cct.lsu.edu>

My email is as private as my paper mail.  I therefore support encrypting
and signing email messages.  Get my PGP key from www.keyserver.net.



-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 186 bytes
Desc: This is a digitally signed message part
Url : http://www.cactuscode.org/pipermail/developers/attachments/20061102/3bcc9357/attachment.bin