TestAllTypes tests CCTK_REAL16, which may not be supported. Since CCTK_REAL16 is used in the interface.ccl, the thorn must then be disabled.

OpenCL is not available on all systems. It could easily be made available on all workstations/laptops, but some older supercomputers won't offer it. There are two open-source OpenCL implementation, but they are alpha quality and don't "just work everywhere".

Finally,  PETSc is not easy to install, and each version has a different interface. PETSc is THE standard, so we need to support it, but making it work (or finding a pre-installed version) is not always easy.

My preferred route would be as Frank suggests: enable them everywhere. We can disable them via Simfactory on machines where they don't work.

-erik



On Sun, May 12, 2013 at 1:51 PM, Frank Löffler <knarf@cct.lsu.edu> wrote:
Hi

Could we, for the development version, enable them for all machines and see what breaks? Maybe we can fix them such that we can unconditionally include them.

Frank

Erik Schnetter <schnetter@cct.lsu.edu> schrieb:


The situation is similar to the OpenCL thorns: They are part of the Einstein Toolkit, part of the release, should be tested (where possible), and should be disabled by default in the thorn list since we assume most people won't be using them. That is inconvenient for testing, but convenient for the end users.

What we really need is a mechanism to easily select a subset of the thorns in the manifest. Simfactory has a mechanism for that, but that is per-machine and may not do what people want. I enable these thorns in my defs.local.ini where they make sense.

-erik



On Sun, May 12, 2013 at 4:48 AM, Ian Hinder <ian.hinder@aei.mpg.de> wrote:

On 11 May 2013, at 21:48, Erik Schnetter <schnetter@cct.lsu.edu> wrote:

I test them regularly. They build fine on many systems (but not all).

The reason we commented them out was that they take a long time to build, while no standard ET thorn uses them. Please include them in the release.

Hi Erik,

Do you mean to included them in the release, but commented-out in the thornlist, or to uncomment them in the thornlist?

If they are in the release, we need to test them as part of the usual process.  This will be awkward if they are commented out.  Everyone will have to remember to uncomment them before running the tests. The automated test system also doesn't know that they should be tested.


-erik


On Sat, May 11, 2013 at 3:21 PM, Frank Loeffler <knarf@cct.lsu.edu> wrote:
On Sat, May 11, 2013 at 01:49:34PM -0400, Erik Schnetter wrote:
> CactusElliptic/EllPETSc
> CactusTest/TestAllTypes
> ExternalLibraries/PETSc
> TAT/TATPETSc

Right. That's because they are commented out in the standard thornlist.
Is this so because they don't build in a lot of places? With them being
commented out they very likely don't receive much testing...

Frank




--
Erik Schnetter <schnetter@cct.lsu.edu>
http://www.perimeterinstitute.ca/personal/eschnetter/
_______________________________________________
Users mailing list
Users@einsteintoolkit.org
http://lists.einsteintoolkit.org/mailman/listinfo/users





--
Erik Schnetter <schnetter@cct.lsu.edu>
http://www.perimeterinstitute.ca/personal/eschnetter/



--
Erik Schnetter <schnetter@cct.lsu.edu>
http://www.perimeterinstitute.ca/personal/eschnetter/