00:21 <hkaiser> K-ballo: nice, looking forward to talk to you tomorrow

00:21 <hkaiser> K-ballo: is 9-9:30am ok?

00:22 <K-ballo> hkaiser: yep, sure

00:23 <hkaiser> k

07:44 <ste||ar-github> [hpx] biddisco pushed 1 new commit to cmake_testing: https://github.com/STEllAR-GROUP/hpx/commit/37152601f478b76625f254379879500ac1aef1f5

07:44 <ste||ar-github> hpx/cmake_testing 3715260 John Biddiscombe: Do not add multi locality tests when Networking is disabled

14:06 <mbremer> heller_: So I got hpx to build, but I needed to add the flag -DMPI_CXX_SKIP_MPICXX=true to my cmake command

16:02 <jayeshbadwaik> Hello, so I wrote in the mail about wanting to work on C++ proposal P0796R3 by Gordon Brown et al. on Supporting Heterogeneous and Distributed Computing Through Affinity. I thought, I would just introduce myself on the irc.

16:07 <hkaiser> hey jayeshbadwaik, welcome

16:08 <zao> Hi there!

16:15 <jayeshbadwaik> Thank you. So, I am looking at the paper, and from what I gather. The main job would be to implement the interface proposed in the paper for discovering the execution resources.

16:16 <jayeshbadwaik> And related tasks, like querying their affinity metric and using those to allocate memory and execute work etc.

16:16 <jayeshbadwaik> Does that sound reasonable?

16:19 <hkaiser> jayeshbadwaik: yes, that's it

16:19 <zao> K-ballo: Couldn't get my build machine to be overly upset about the HPX codebase with CUDA, just some hello_compute example that failed to build.

16:20 <hkaiser> zao: K-ballo is here with us now, he might be interested in looking into the CUDA builds

16:21 <hkaiser> jayeshbadwaik: you might want to use hwloc to get at the information to be exposed

16:22 <hkaiser> so reading about the hwloc API might be advisable as well

16:23 <jayeshbadwaik> hkaiser: I see. Sure, I will do that. Right now, I am trying to map out my approach to the task. So, the paper proposes two interfaces: 1. A low granularity interface which is aimed at less-knowledgeable users and a high granularity interface which exposes a lot more. In my opinion, it would be good to first concentrate on getting the high granularity interface working. And then implement low granularity later.

16:24 <hkaiser> that's the correct way to do things, however it might be easier to start with the low granularity API, the do the high granularity one and then redo the further on top of the latter ;-)

16:24 <hkaiser> but that's for you to decide

16:25 <jayeshbadwaik> Hmm. :-) Cool, thanks. So, I will read up more on hwloc API and let you know if I have more questions.

16:26 <hkaiser> jayeshbadwaik: you also might want to look at what we have today in HPX on top of hwloc

16:26 <jayeshbadwaik> I will.

16:34 <K-ballo> zao: using master? could you try the `no-return-type-deduction` branch and see if it is any worse?

16:34 <K-ballo> although I have to figure out those return type deduction induced failures anyhow

16:36 <zao> K-ballo: I linked gists last night with current master and then-current tip of n-r-t-d.

16:36 <zao> Building with C++11.

16:36 <K-ballo> I'll search the archives

16:37 <zao> 18:59 <zao> Current master in C++11 mode w/ GCC 7.3.0 and CUDA 9.2 - https://gist.github.com/b158a8dc6a1c3fcf16cfb381494d0b63

16:37 <zao> 19:05 <zao> There's some errors in there, I built -k0 to get it to cover the whole codebase.

16:37 <zao> 19:50 <zao> Same with no-return-type-deduction branch - https://gist.github.com/a2a0fc8b40b05d7b79a1c0039ecf6de8

16:37 <K-ballo> thanks

16:45 <zao> The machine has no NV GPU so I can't actually test any code, but the builds seem simple enough.

16:46 <K-ballo> the build doesn't fail (more), that's important

17:22 <ste||ar-github> [hpx] msimberg pushed 1 new commit to release: https://github.com/STEllAR-GROUP/hpx/commit/1733acbf13666d644e647ac56b2682d5a2bec188

17:22 <ste||ar-github> hpx/release 1733acb Mikael Simberg: Merge remote-tracking branch 'origin/master' into release

18:08 <ste||ar-github> [hpx] StellarBot pushed 1 new commit to gh-pages: https://github.com/STEllAR-GROUP/hpx/commit/49d9071b0b803c309df5c75eb38b5abcac80c8c6

18:08 <ste||ar-github> hpx/gh-pages 49d9071 StellarBot: Updating Sphinx docs

18:34 <ste||ar-github> [hpx] K-ballo force-pushed no-return-type-deduction from c05b50e to 5cd8dc1: https://github.com/STEllAR-GROUP/hpx/commits/no-return-type-deduction

18:34 <ste||ar-github> hpx/no-return-type-deduction f5e714b Agustin K-ballo Berge: (trigger incomplete type failures)

18:34 <ste||ar-github> hpx/no-return-type-deduction 5cd8dc1 Agustin K-ballo Berge: (draft)

18:53 <ste||ar-github> [hpx] K-ballo force-pushed no-return-type-deduction from 5cd8dc1 to eca9d27: https://github.com/STEllAR-GROUP/hpx/commits/no-return-type-deduction

18:53 <ste||ar-github> hpx/no-return-type-deduction eca9d27 Agustin K-ballo Berge: (draft)

19:15 <mbremer> Random question. Can I have an HPX component of a pure virtual class?

19:38 <heller_> yes

19:38 <heller_> look for derived component

19:48 <mbremer> heller_: managed_component? Sweet I'll look into it. The tests are pretty helpful for what I want to do. I'll start there

19:48 <heller_> they should work, yes

19:49 <ste||ar-github> [hpx] K-ballo force-pushed no-return-type-deduction from eca9d27 to 0c500b7: https://github.com/STEllAR-GROUP/hpx/commits/no-return-type-deduction

19:49 <ste||ar-github> hpx/no-return-type-deduction 0c500b7 Agustin K-ballo Berge: (draft)

19:51 <ste||ar-github> [hpx] biddisco pushed 2 new commits to guided_executor: https://github.com/STEllAR-GROUP/hpx/compare/9396a2533b9f...99c6faca4b7f

19:51 <ste||ar-github> hpx/guided_executor 38ce757 John Biddiscombe: Fix a problem with recent cmake+MPI MPIEXEC vars

19:51 <ste||ar-github> hpx/guided_executor 99c6fac John Biddiscombe: Changes require for C++11 compilation

20:37 <ste||ar-github> [hpx] biddisco force-pushed guided_executor from 99c6fac to 1852e57: https://github.com/STEllAR-GROUP/hpx/commits/guided_executor

20:37 <ste||ar-github> hpx/guided_executor 1852e57 John Biddiscombe: Changes require for C++11 compilation

22:02 <ste||ar-github> [hpx] K-ballo force-pushed no-return-type-deduction from 0c500b7 to 67da91b: https://github.com/STEllAR-GROUP/hpx/commits/no-return-type-deduction

22:02 <ste||ar-github> hpx/no-return-type-deduction 67da91b Agustin K-ballo Berge: Replace executor traits return type deduction (keep non-SFINAE)

22:08 <ste||ar-github> [hpx] biddisco force-pushed guided_executor from 1852e57 to c6207ec: https://github.com/STEllAR-GROUP/hpx/commits/guided_executor

22:08 <ste||ar-github> hpx/guided_executor c6207ec John Biddiscombe: Changes require for C++11 compilation

22:32 <mbremer> Follow-up question: I'm getting ambiguous lookups for the managed components. Here is the compiler error https://gist.github.com/bremerm31/65e644aa5b0720464956dc7dd3f3bf60

22:32 <mbremer> Is this something I'm doing wrong

22:33 <mbremer> ?

22:51 <hkaiser> mbremer: show me your component definition

22:51 <hkaiser> mainly what is it derived from

22:52 <mbremer> I'm trying to derive it from an abstract base class. I'm mainly following the tests in tests/unit/inheritance_2_classes_abstract.cpp

22:53 <hkaiser> mbremer: nod

22:53 <mbremer> Let me commit and push, and I can show you code.

22:57 <mbremer> hkaiser: Here is the base class https://github.com/bremerm31/dgswemv2/blob/virtual_problem_class/source/simulation/hpx/sim_unit_hpx_base.hpp

22:58 <mbremer> and here is the Derived class https://github.com/bremerm31/dgswemv2/blob/virtual_problem_class/source/simulation/hpx/sim_unit_hpx.hpp

22:58 <mbremer> Is there anything in particular you are looking for?

22:58 <hkaiser> mbremer: and that one gives you the error?

22:59 <mbremer> so my understanding is that some typedef in the derived component is ambiguous

23:00 <hkaiser> only if you derive from two base classes

23:00 <hkaiser> the abstract base class test does it correctly

23:01 <mbremer> Oh I see, you manually specify the wrapping_type there

23:01 <mbremer> and the other typedefs

23:01 <hkaiser> mbremer: the test explicitly disambiguates things: https://github.com/STEllAR-GROUP/hpx/blob/master/tests/unit/component/inheritance_3_classes_1_abstract.cpp#L73-L78

23:02 <mbremer> ! Let me try that, but I would expect that to do the trick. Thanks @hkaiser