[Dune] Is anyone using AlbertaGrid with -DDUNE_ALBERTA_CACHE_COORDINATES=0 ?

[Martin Nolte]

Hi Jö,

up to now, I have only read your proposal once (lack of time). There seem to 
be some subtle points in it and I have to think through more thoroughly. I'd 
suggest to talk on sunday.

For Alberta, I think the crucial point is that entity and geometry objects may 
only be read by another thread. The subtle point is that AlbertaGrid uses 
shallow copies (not thread-safe up to now). So, surprisingly, you cannot count 
copying a read-only operation. But I think you already anticipated this problem.

As ALBERTA internals might also be involved in this discussion, I would 
suggest a small discussion between Claus Heine, you and me (and everyone 
interested).

Best,

Martin

On 09/24/2015 11:39 AM, Jö Fahlke wrote:
> Am Thu, 24. Sep 2015, 08:35:04 +0200 schrieb Martin Nolte:
>> Date: Thu, 24 Sep 2015 08:35:04 +0200
>> From: Martin Nolte <nolte at mathematik.uni-freiburg.de>
>> To: dune at dune-project.org
>> Subject: Re: [Dune] Is anyone using AlbertaGrid with
>>   -DDUNE_ALBERTA_CACHE_COORDINATES=0 ?
>>
>> Hi Jö,
>>
>> actually, I wanted to switch the default once ALBERTA 2.0 is gone. As of
>> version 3,0, ALBERTA can always pass the coordinates to our entities (even
>> though they might not be needed).
>>
>>  From what I remember, however, the code should be view thread-safe (unless
>> you migrate entity or geometry objects between threads. But I must admit
>> that I have not fully understood your proposal, yet.
>
> Well, it allows migrating objects between threads, and even using them from
> multiple threads concurrently, as long as these uses are not modifications.  I
> guess I'll have a closer look at the Alberta implementation, then.
>
> Is there anything in particular about the proposal, which you would like
> clarification on, or do you want to postpone that to the meeting?
>
> Regards,
> Jö.
>

-- 
Dr. Martin Nolte <nolte at mathematik.uni-freiburg.de>

Universität Freiburg                                   phone: +49-761-203-5630
Abteilung für angewandte Mathematik                    fax:   +49-761-203-5632
Hermann-Herder-Straße 10
79104 Freiburg, Germany