[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Getfem-users] [[fallthrough]]

From: Andriy Andreykiv
Subject: Re: [Getfem-users] [[fallthrough]]
Date: Sun, 10 Mar 2019 15:07:14 +0100

Dear Konstantinos,

Thank you for the heads up. I will monitor your commits. Just a small remark. The only thing we are always worried about is thread safety. On our side we always assemble in parallel with OpenMP. So if you make any modifications to ga_workspace or model please be careful with static or shared variables. 

Best regards, Andriy 

On Sun, 10 Mar 2019, 10:02 am Konstantinos Poulios, <address@hidden> wrote:
Dear Andriy,

Sorry for breaking your use case. I wasn't aware of that. In the coming days I will progressively commit some more restructuring of core parts of GetFEM such as the model and ga_workspace classes. These changes are necessary for my implementation of internal variable condensation. Please keep an eye on what is committed to the master branch and let me know if some commit breaks any of your code in a non fixable way. Of course to avoid breakage in general we should extend our test harness to cover all use cases.

Best regards

On Fri, Mar 8, 2019 at 3:11 PM Andriy Andreykiv <address@hidden> wrote:
... additionally,

I moved model_pb template back to the header, as it's being used in our code
Best regards,

On Fri, 8 Mar 2019 at 14:22, Andriy Andreykiv <address@hidden> wrote:
Dear Tetsuo Koyama,

A few weeks ago you introduced [[fallthrough]] in a few places of Getfem.
I completely agree with your intention but the problem with this feature is that
it's supported only as of C++17. Many users including us are not yet capable to
upgrade to the compilers that support this standard. Default compilation of getfem
is currently at C++ 11, but may be GCC supports it (I don't know).

May I ask you to replace [[fallthrough]] with some kind of macro that is only triggered
when standard 17 is being used?

Thank you,

reply via email to

[Prev in Thread] Current Thread [Next in Thread]