Re: [hpx-users] Debugging bad locality number on MPI?
Heya, On 11:50 Fri 05 Jul , Hartmut Kaiser wrote: > Could you please list your MPI env variables here? Sure, here is a dump: https://gist.github.com/gentryx/c94c135862b1c894b59b2f333c9198bf > The number of localities should be picked up from them before the > parcelport is initialized. Something goes wrong there, so it assumes > that networking should not be enabled at all... Huh, is that portable? I thought the MPI standard requires you to call MPI functions to discover them. Anyway, I see that in plugins/parcelport/mpi/mpi_environment.cpp detect_mpi_environment() will look for OMPI_COMM_WORLD_SIZE, but this is only running if the MPI parcelport is to be instantiated, that's after is_networking_enabled() is run. Cheers -Andi -- ========== Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] Debugging bad locality number on MPI?
Thanks Thomas! The MPI environment variables are set up correctly. I dug a bit deeper and figured out that HPX isn't even enabling networking. In hpx/src/util/runtime_configuration.cpp the function enable_networking() only enables networking if one of the following conditions is true: a) the number of localities is > 1 b) the node number (is this the rank?) is > 0 c) the number of expected localities is != 0 d) the runtime mode is not console In my case the values are a) 1 b) -1 c) 0 d) console These seem to be correct, since HPX can't know the number of localities prior to initializing the MPI parcelport. (enable_networking() is run when the parcelhandler is created, but before the parcelports are instantiated. I'll be honest: I don't quite understand the logic there, but if I change the code to enable networking in console mode, or if I enable networking if the node number equals -1, then it works. Looks like the this was introduced with https://github.com/STEllAR-GROUP/hpx/commit/ffb8470e6a1143e9e1c95c39eff58eec322148d3#diff-86850321552971332a5de979a34bd259 Thoughts? Thanks! -Andi On 10:45 Fri 05 Jul , Thomas Heller wrote: > The relevant parts in the codebase are here: > Setting the env to check via cmake: > https://github.com/STEllAR-GROUP/hpx/blob/master/plugins/parcelport/mpi/CMakeLists.txt#L42 > Detecting if we run inside a MPI environment: > https://github.com/STEllAR-GROUP/hpx/blob/master/plugins/parcelport/mpi/mpi_environment.cpp#L30-L52 > > On Fri, Jul 5, 2019 at 10:41 AM Thomas Heller wrote: > > > The initialization of the MPI parcelport is done by checking environment > > variables set by the most common MPI implementations. Which MPI > > implementation do you use? Can you attach the output of `mpirun env` maybe? > > > > Andreas Schäfer schrieb am Fr., 5. Juli 2019 08:58: > > > >> > >> On 01:11 Fri 05 Jul , Marcin Copik wrote: > >> > I've seen such issue in pure MPI applications. The usual reason is > >> > using an mpirun/mpiexec provided by an implementation different than > >> > the one used for linking. Checking for a mismatch there might help. > >> > >> There is only one MPI version installed on that machine. Also, running > >> a simple MPI hello world works as expected. My assumption is that the > >> MPI parcelport is not initialized correctly. Which part of the code > >> loads/initialized all the parcelports? > >> > >> Thanks > >> -Andi > >> > >> > >> -- > >> == > >> Andreas Schäfer > >> > >> HPC and Supercomputing for Computer Simulations > >> LibGeoDecomp project lead, http://www.libgeodecomp.org > >> > >> PGP/GPG key via keyserver > >> > >> I'm an SRE @ Google, this is a private account though. > >> All mails are my own and not Google's. > >> == > >> > >> (\___/) > >> (+'.'+) > >> (")_(") > >> This is Bunny. Copy and paste Bunny into your > >> signature to help him gain world domination! > >> ___________ > >> hpx-users mailing list > >> hpx-users@stellar.cct.lsu.edu > >> https://mail.cct.lsu.edu/mailman/listinfo/hpx-users > >> > > > ___ > hpx-users mailing list > hpx-users@stellar.cct.lsu.edu > https://mail.cct.lsu.edu/mailman/listinfo/hpx-users -- == Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] Debugging bad locality number on MPI?
On 01:11 Fri 05 Jul , Marcin Copik wrote: > I've seen such issue in pure MPI applications. The usual reason is > using an mpirun/mpiexec provided by an implementation different than > the one used for linking. Checking for a mismatch there might help. There is only one MPI version installed on that machine. Also, running a simple MPI hello world works as expected. My assumption is that the MPI parcelport is not initialized correctly. Which part of the code loads/initialized all the parcelports? Thanks -Andi -- ====== Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] Google Summer of Code
On 02:00 Wed 25 Apr , Gabriel Laberge wrote: > Right now I'm focusing on finishing my last exams, but afterwards I > intend to begin my project. Cool, good luck for your exams! Cheers -Andi -- == Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] CUDA with HPX
Ivan, could you retest this with CUDA 9.1? Version 7.5 is really old and I don't think it works well with modern C++. I don't think you have to re-install the whole OS, you could alternatively just download CUDA and install it manually. Thanks! -Andreas On 14:22 Tue 27 Mar , Ivan Kostov wrote: > Hi, > > I am trying to understand how HPX Compute works with CUDA > and wanted to compile the > examples/compute/cuda/data_copy.cu example. Unfortunately, > I'm getting the following error: > > /usr/include/c++/5/bits/stl_iterator_base_types.h(154): > error: class > "std::unique_ptr<hpx::parcelset::locality::impl_base, > std::default_delete>" > has no member "iterator_category" >detected during: > instantiation of class > "std::__iterator_traits<_Iterator, void> [with > _Iterator=std::unique_ptr<hpx::parcelset::locality::impl_base, > std::default_delete>]" > (163): here > instantiation of class > "std::iterator_traits<_Iterator> [with > _Iterator=std::unique_ptr<hpx::parcelset::locality::impl_base, > std::default_delete>]" > /usr/local/include/hpx/traits/is_iterator.hpp(43): here > instantiation of class > "hpx::traits::detail::is_iterator [with > T=std::unique_ptr<hpx::parcelset::locality::impl_base, > std::default_delete>]" > /usr/local/include/hpx/traits/is_iterator.hpp(49): here > instantiation of class > "hpx::traits::is_iterator<Iter, Enable> [with Iter=const > std::unique_ptr<hpx::parcelset::locality::impl_base, > std::default_delete> > &, Enable=void]" > /usr/local/include/hpx/runtime/parcelset/locality.hpp(150): > here > > /usr/include/c++/5/bits/stl_iterator_base_types.h(155): > error: class > "std::unique_ptr<hpx::parcelset::locality::impl_base, > std::default_delete>" > has no member "value_type" >detected during: > instantiation of class > "std::__iterator_traits<_Iterator, void> [with > _Iterator=std::unique_ptr<hpx::parcelset::locality::impl_base, > std::default_delete>]" > (163): here > instantiation of class > "std::iterator_traits<_Iterator> [with > _Iterator=std::unique_ptr<hpx::parcelset::locality::impl_base, > std::default_delete>]" > /usr/local/include/hpx/traits/is_iterator.hpp(43): here > instantiation of class > "hpx::traits::detail::is_iterator [with > T=std::unique_ptr<hpx::parcelset::locality::impl_base, > std::default_delete>]" > /usr/local/include/hpx/traits/is_iterator.hpp(49): here > instantiation of class > "hpx::traits::is_iterator<Iter, Enable> [with Iter=const > std::unique_ptr<hpx::parcelset::locality::impl_base, > std::default_delete> > &, Enable=void]" > /usr/local/include/hpx/runtime/parcelset/locality.hpp(150): > here > > /usr/include/c++/5/bits/stl_iterator_base_types.h(156): > error: class > "std::unique_ptr<hpx::parcelset::locality::impl_base, > std::default_delete>" > has no member "difference_type" > > > and it continues like this for a while. > > I am running an Ubuntu 16.04, gcc 5.4.0., cuda 7.5.17 > > this is the cmake file i used to compile the example > > cmake_minimum_required(VERSION 3.3) > project(hpxvr CXX) > > set(CMAKE_CXX_STANDARD 14) > set(CMAKE_CXX_STANDARD_REQUIRED ON) > > set(HPX_IGNORE_COMPILER_COMPATIBILITY ON) > > find_package(HPX REQUIRED) > > include_directories(${HPX_INCLUDE_DIR}) > > set( CMAKE_CXX_FLAGS "-D_MWAITXINTRIN_H_INCLUDED > -D_FORCE_INLINES -D__STRICT_ANSI__") > > add_hpx_executable(cuda_copy_expl > ESSENTIAL > SOURCES data_copy.cu > ) > > I used to have another two error messages > > /usr/lib/gcc/x86_64-linux-gnu/5/include/mwaitxintrin.h(36): > error: identifier "__builtin_ia32_monitorx" is undefined > > /usr/lib/gcc/x86_64-linux-gnu/5/include/mwaitxintrin.h(42): > error: identifier "__builtin_ia32_mwaitx" is undefined > > which were solved using this workaround - > https://github.com/NVIDIA/nccl/issues/29 > > As a last case solution we could upgrade to Ubuntu 17 and > hope that the problem resolves itself with newer CUDA > versions, but would rather remain on Ubuntu 16. Do you > have any solution ideas that could help? > > Best regards, > Ivan > > > > ___ > hpx-users mailing list > hpx-users@stellar.cct.lsu.edu > https://mail.c
Re: [hpx-users] GSoC2018 Newtonian Physics Sandbox
I just went over the changed parts of your proposal. LGTM. :-) You're probably planning on doing this anyway, but I'll say it anyway: you should go ahead and remove the sections that are stricken out and resolve all comments that have been addressed, so that the doc looks nicer. Thanks! -Andi On 18:43 Mon 26 Mar , Evgeny wrote: > Hey, > > The last day proposal can be submitted. I have already made the > cleaned version in PDF, if there is any more clarification, you can write. > > Sincerely, > Evgeny Dedov -- ========== Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] GSoC2018 Newtonian Physics Sandbox
Heya, On 01:52 Sat 24 Mar , Evgeny wrote: > I have a question about your bouncing spheres example. You said, that > spheres don't actually see a grid, only neighboring spheres, but how > neighborhood choses actually, I caught myself thinking, when I went deep > enough into the source code of the library itself, but then realized that > obviously it should be set by some api, but can't found which. very good question. So, in the example code you've probably seen that main() instantiates a SerialSimulator [1]. The SerialSimulator actually operates on a 3d, regular grid. The reason that the spheres only see neighboring spheres (and boundary elements) is because Container provides adapters that translate the neighborhood of grid cells into a neighborhood of particles (and boundaries). Container inherits from BaseContainer[2]. This class is generated via DECLARE_MULTI_CONTAINER_CELL() and does some behind the scenes magic. Here is a quick rundown of what happens during simulation: 1. SerialSimulator will call update on all instances of Container in the 3D grid. 2. Container::update() is actually BaseContainer::update(), because inheritance. It receives a neighborhood object with can be addressed via relative coordinates (e.g. Coord<3>(-1, 0, 0)) to retrieve neighboring Container objects. 3. BaseContainer::update() now iterates over all elements in its members "spheres" and "boundaries" and calls Sphere::update() and Boundary::update() respectively. But instead of passing the neighborhood object it got from the SerialSimulator, it will pass on a new neighborhood object[3], a MultiNeighborhoodAdapter[4]. This has two members, called "spheres" and "boundaries", which are generated in the expansion of the macro DECLARE_MULTI_CONTAINER_CELL(). 4. These members are actually of type NeighborhoodIteratorHelpers::Adapter[5]. This object does the translation from the neighborhood that is defined on the regular grid towards iterating over spheres (or boundaries). So yeah, definitely a lot going on under the hood. :-) Did this answer your question? Feel free to ask more. Cheers -Andi [1] https://github.com/STEllAR-GROUP/libgeodecomp/blob/master/src/examples/bouncingspheres/main.cpp#L384 [2] https://github.com/STEllAR-GROUP/libgeodecomp/blob/master/src/examples/bouncingspheres/main.cpp#L165 [3] https://github.com/STEllAR-GROUP/libgeodecomp/blob/master/src/libgeodecomp/storage/multicontainercell.h#L157 [4] https://github.com/STEllAR-GROUP/libgeodecomp/blob/master/src/libgeodecomp/storage/multicontainercell.h#L109 [5] https://github.com/STEllAR-GROUP/libgeodecomp/blob/master/src/libgeodecomp/storage/neighborhooditerator.h#L19 -- ====== Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] hpx-users Digest, Vol 58, Issue 27
On 16:49 Wed 21 Mar , Hartmut Kaiser wrote: > Well Andreas Schaefer (gentryx) seems to be following this list, so he might > respond to your inquiry. As I wrote in the other mail: yes, the project is still relevant. :-) -- == Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] GSoC2018 Newtonian Physics Sandbox
Hey, On 00:40 Thu 22 Mar , Evgeny wrote: > Hello! I spent time discovering examples and tests provided. All built and > runs ok. that's great! > So if you are more attainable in mail, It is ok, I will ask you > using email, if you offline in IRC. Yes, please send your questions via mail. Thanks -Andi -- ========== Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] GSoC 2018
Hey Alanas, thanks for getting in touch. On 19:45 Fri 16 Mar , Alanas Plascinskas wrote: > Hey, > > I wanted to ask about the availability of your projects for GSoC 2018. > > I would mainly be interested in these projects: All to All Communications, > Conflict (Range-Based) Locks and Newtonian Physics Sandbox but I"m open to > any suggestions as well. The Newtonian Physics Sandbox project is available until someone solves it. :-) > Have any of these received any serious proposals and if so do you have any > other projects that would be important for you and are still open? I haven't seen any proposal this year yet. In my experience they usually come in hours before the deadline. That said, acceptance is, as with probably all GSoC projects, conditional to a vote from all mentors. So having no competition doesn't mean a free ticket. The proposal still needs to be good. Let me know if I can answer any questions on it. Thanks -Andi -- ========== Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] gsoc2018
Hey Sudhir, On 14:16 Fri 16 Mar , SUDHIR KUMAR SUMAN wrote: > Hi > I am SUDHIR KUMAR SUMAN. I am second year Btech student at IIT Bombay > ,INDIA. > I have good experience of in programming in C++.I use write alogrithm to > solve problem related to physical world like collision of objects,rotation > of objection,etc . > I would like to work on project "NEWTONIAN PHYSICS SANDBOX". > I have done some short term project related to the collision of n object > in one dimensional space where the algorithm will give the velocity at and > position of objects after n collisions. interesting. Was that an event-based algorithm or was it force-based? Which interaction model did you use? > If this is considered worthwhile by potential mentors, I would get > started on writing a proposal right away. Yes, I'd still like to mentor such a project. Ultimately we'll do a vote on all projects (we usually get way more applications than slots for students), so your proposal would also need the support from other mentors from the Ste||ar Group. A good starting point, as Patrick has already pointed out, is to check out the trunk of LibGeoDecomp, get it compiled and run the tests. Thanks -Andi -- ========== Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] GSoC2018 Newtonian Physics Sandbox
Hey, just found your initial mail. :-) On 21:11 Mon 12 Mar , Evgeny wrote: > Hello, > I found intresting contributing in HPX in this GSoC > I'm well-versed in C++ and have some parallel programming expirience (MPI, > OpenMP) > > One of projects atreccted my attention most is legacy project Newtonian > Physics Sandbox, I'm in the second year of the IT department in Novosibirsk > State University, before that I studied at the physics and mathematics > school and attended a course "physical processes modelling" in C ++, where > during the training was modelling many processes, such as lissajous curves, > motion in a central force field, celestial bodies interaction, tunnel diode > etc. Not all of these related to Newtonian physics, I mentioned them for a > change. Cool, that sounds like a background that would make you a good fit for this project. > Also I haven't found any of beginnings of project, so as I understand, it > does not exist yet, am I right? > I gradually began to get acquainted with LibGeoDecomp tool, and think I got > the idea of it. If there are some other materials or advices for acquaintance, > I will be grateful. Correct, no such toolbox has been written. LibGeoDecomp has been used in similar settings, e.g. for smoothed particle hydrodynamics or for simulations of granular gasses, but these were always rather specialized codes and nothing generic or really reusable. I thing the first step, as it is so often, is actually to check out the trunk from https://github.com/STEllAR-GROUP/libgeodecomp and see if you can compile the library and run the tests. Cheers -Andi -- ====== Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] GSoC'18 LibGeoDecomp Support
Hey, On 16:23 Wed 21 Mar , John Biddiscombe wrote: > I think the libgeocomp project must have been deprecated by now. The > chap who was in charge of it has since left the HPX group and is no > longer involved. I didn't work on LibGeoDecomp while I was ramping up at Google. It's not deprecated though. ATM I'm working on getting it to work with HPX trunk again. Cheers -Andi ps: John, ping me if you ever pass through Zürich. :-) -- ====== Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] GSoC'18 LibGeoDecomp Support
Hey Evgeny, I'm the maintainer of LibGeoDecomp and LibFlatArray. On 22:17 Wed 21 Mar , Evgeny wrote: > Hello, who can I ask about some LibGeoDecomp specific questions? I quite > good understand how things are there, but some times documentation and > source code do not provide enough information to quickly understand. I > tried to ask in IRC few times, but seems like there I can get some HPX > support only, and I can't got answer. Yeah, I'm rarely on IRC these days, sorry about that. Documentation is sparse, I believe the best source for that is usually the unit tests. How can I help you, what do you need to know? Thanks -Andi -- ====== Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] GSOC questions about project
Hey Александр, On 12:10 Mon 13 Mar , Александр Юрьевич Кондратьев wrote: > I have a question about the project "Extension LibGeoDecomp::Region as an > alternative adjacency container to BGL". > I correctly understand that the purpose of the project is to implement the > class Region in BGL without external dependencies to LibGeoDecomp? no, the goal is to do the opposite: utilize LibGeoDecomp::Region insilde BGL as an adjacency container. The motivation behind this is that LibGeoDecomp::Region is highly efficient for storing coordinate sets. And the edges of a directed graph with an integer vertex set can be considered 2D coordinates. Does that make sense? Thanks! -Andreas -- ========== Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though. All mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] GSOC 2017 - Newtonian Physics Sandbox Project
CCing HPX Users. Let's keep this discussion public so others can chime in. Thanks! -Andreas On 09:32 Fri 10 Mar , Andreas Schäfer wrote: > Hey Neeladree, > > On 15:19 Thu 09 Mar , Neeladree Choudhury wrote: > > I'm Neeladree from National Institute of Technology, Warangal in India. I > > am interested in the Newtonian Physics Sandbox Project . > > Cool! > > > I'm quite proficient in writing code in C++. I have participated in 2 ACM > > ICPC Regionals in 2016. > > > > I have also studied Physics courses at the higher secondary level and am > > familiar with concepts of Newtonian Physics. > > Wow, both should give you a huge advantage. > > > I would like to know more about the kind and extent of physical modelling > > required in this project. > > Actually the physics part doesn't have to be all too complex. I'd be > good with a linear dashpot model and an Euler integrator. In my > experience the major challenge here is to get comfortable with the > tooling we'll use. Have you tried compiling HPX yet? Also, what OS is your > primary work machine running? Windows, Linux, macOS? > > Cheers > -Andreas > > > -- > == > Andreas Schäfer > > PGP/GPG key via keyserver > http://www.libgeodecomp.org > == > > (\___/) > (+'.'+) > (")_(") > This is Bunny. Copy and paste Bunny into your > signature to help him gain world domination! -- == Andreas Schäfer HPC and Supercomputing for Computer Simulations LibGeoDecomp project lead, http://www.libgeodecomp.org PGP/GPG key via keyserver I'm an SRE @ Google, this is a private account though, all mails are my own and not Google's. == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] [SOLVED] Attempting to do sparse triangular solve
Hey Riccardo, On 11:26 Sat 10 Dec , Riccardo Rossi wrote: > i finally managed to make my futurized sparse triangular solver to work. > over the next days i'll try it for performance, but for this i'll need to > develop a blocked version and some real CSR matrix. FWIW the Matrix Market is a source for sparse matrices that's well published and hence lends itself for comparison with the state of the art: http://math.nist.gov/MatrixMarket/formats.html Is there a specific reason why you're using CSR? AFAIK other formats are more efficient for using sparse matrices in computational kernels. Cheers -Andreas -- ====== Andreas Schäfer HPC and Supercomputing Institute for Multiscale Simulation Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-20866 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
[hpx-users] HPX+LGD test failures
Quick question: I'm getting the error message I've put into the gist below. It's triggered when I trigger a remote action, e.g. hpx::apply(ReceiverType3::receiveAction(), leftID, 30, sendToLeft); I'm currently on LGD master and HPX 2d93024be0aa27d320ae4200f58b1f24467ba5e9, but I've seen the same with HPX master. The same test is running on the other machines which are running my CI tests, but it's failing on my notebook. Any ideas what might be wrong here? https://gist.github.com/gentryx/1d18d0d735d2ddd52f9fabdac4972e0f Thanks -Andi -- == Andreas Schäfer HPC and Supercomputing Institute for Multiscale Simulation Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-20866 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] Iterative construction of dataflow graph
On 14:32 Wed 24 Aug , Andreas Schäfer wrote: > You can limit the tests to be run in LibGeoDecomp by setting > LIMIT_FLAGS in CMake: D'oh, typo fixed: cd libgeodecomp4/build/ cmake -DLIMIT_TESTS=parallelization.test.parallel_hpx_4 .. make tests make check -- == Andreas Schäfer HPC and Supercomputing Institute for Multiscale Simulation Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-20866 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] Iterative construction of dataflow graph
On 07:21 Wed 24 Aug , Hartmut Kaiser wrote: > > > the HPXDataflowSimulator doesn't use async() anywhere. It's completely > > dataflow-based. > > Ok, something is not quite right, then. Can you provide me with your small > example? Will that run on a single node (for simplicity)? I guess the easiest way would be to simply build LibGeoDecomp with HPX support and uncommenting the test testManySteps() in libgeodecomp/src/parallelization/test/parallel_hpx_4_1/hpxdataflowsimulatortest.h. If you set the chunk size in the test to a high value, then it should exhaust that system's memory: HPXDataflowSimulator sim(initializer, "testManySteps", 30, true, 30); You can limit the tests to be run in LibGeoDecomp by setting LIMIT_FLAGS in CMake: cd libgeodecomp4/build/ cmake -DLIMIT_TESTS=parallelization.test_parallel_hpx_4 .. make tests make check Cheers -Andi -- ========== Andreas Schäfer HPC and Supercomputing Institute for Multiscale Simulation Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-20866 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] Iterative construction of dataflow graph
Hartmut- the HPXDataflowSimulator doesn't use async() anywhere. It's completely dataflow-based. Cheers -Andi On 06:44 Wed 24 Aug , Hartmut Kaiser wrote: > Andy, > > I probably forgot to mention that not only the dataflow() calls (those we > could even get away without it), but also all async() should be forked. > > Regards Hartmut > --- > http://boost-spirit.com > http://stellar.cct.lsu.edu > > > > -Original Message- > > From: Andreas Schäfer [mailto:gent...@gmx.de] > > Sent: Wednesday, August 24, 2016 5:37 AM > > To: hartmut.kai...@gmail.com; hpx-users@stellar.cct.lsu.edu > > Cc: 'Zach Byerly' <zbye...@gmail.com> > > Subject: Re: [hpx-users] Iterative construction of dataflow graph > > > > Heya, > > > > quick update: I was able to reproduce the effects described by Zach > > with a simple test model[1], which breaks down (i.e. sends my test > > machine into swapping with just 60 simulation entities) at 100k time > > steps. > > > > On 07:41 Fri 19 Aug , Hartmut Kaiser wrote: > > > To make a long story short, I'd suggest to change all dataflow(...) > > function > > > calls in DGSWEM/HPX with dataflow(hpx::launch::fork, ...) and see if > > this > > > helps the issue. It actually should - it would be nice to see whether > > theory > > > and practice are actually the same for a change ;) > > > > I did measure hpx::launch::async and hpx::launch::fork for less than > > 100k time steps (results attached). async broke down at ~75k time > > steps, fork did mitigate this effect up to ~95k time steps but broke > > down as well. > > > > I then implemented a simple chunked approach with a barrier every > > "chunkSize" time steps. This is terrible as it forces all cores to > > drain prior to generating more work, but my tries to overlap dataflow > > generation with work were unsuccessful (I spent 2h on this but only > > got deadlocks. I'd be happy to merge a PR that would this issue). > > chunkSize is configurable, though, so the actual effect shouldn't be > > too bad. With this change I can easily go up to 300k time steps > > without seeing any degradation of the time taken per time step. > > > > Cheers > > -Andi > > > > [1] > > https://github.com/gentryx/libgeodecomp/commit/c7fa53f4462b3562a189a2f5212 > > 21e6ddb22cafc > > [2] > > https://github.com/gentryx/libgeodecomp/commit/6a4a6a747484476b291d00a23fe > > 515ffcd49b8d2 > > > > > > -- > > == > > Andreas Schäfer > > HPC and Supercomputing > > Institute for Multiscale Simulation > > Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany > > +49 9131 85-20866 > > PGP/GPG key via keyserver > > http://www.libgeodecomp.org > > == > > > > (\___/) > > (+'.'+) > > (")_(") > > This is Bunny. Copy and paste Bunny into your > > signature to help him gain world domination! > > > -- == Andreas Schäfer HPC and Supercomputing Institute for Multiscale Simulation Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-20866 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] Build problems with 73c6b8aecaa9433a6fe523aa579b0d252660726e
That did the trick. Thanks! -Andi On 09:49 Wed 29 Jun , Hartmut Kaiser wrote: > Andy, > > Yes, master is currently broken, working on it... (well, some examples/tests > don't build) > Please use the 'release' branch until after the release in two weeks, we'll > keep that 100% stable. > > Regards Hartmut > --- > http://boost-spirit.com > http://stellar.cct.lsu.edu > > > > -Original Message- > > From: hpx-users-boun...@stellar.cct.lsu.edu [mailto:hpx-users- > > boun...@stellar.cct.lsu.edu] On Behalf Of Andreas Schäfer > > Sent: Wednesday, June 29, 2016 6:51 AM > > To: HPX Users <hpx-users@stellar.cct.lsu.edu> > > Subject: [hpx-users] Build problems with > > 73c6b8aecaa9433a6fe523aa579b0d252660726e > > > > Heya, > > > > I'm trying to build HPX with the following options: > > > >cmake -DCMAKE_INSTALL_PREFIX=/home/gentryx/local_install/ - > > DCMAKE_CXX_COMPILER=g++-4.9.3 -DHPX_WITH_PARCELPORT_MPI=true - > > DHPX_WITH_PARCELPORT_TCP=true -DHPX_WITH_CXX14=false - > > DCMAKE_BUILD_TYPE=Release - > > DHPX_WITH_CXX11_EXTENDED_FRIEND_DECLARATIONS=false .. > > > > ...which have worked for previous builds. However, the build fails > > with the attached error message. Any ideas what might be going wrong? > > > > Thanks! > > -Andreas > > > > > > -- > > == > > Andreas Schäfer > > HPC and Supercomputing > > Institute for Multiscale Simulation > > Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany > > +49 9131 85-20866 > > PGP/GPG key via keyserver > > http://www.libgeodecomp.org > > == > > > > (\___/) > > (+'.'+) > > (")_(") > > This is Bunny. Copy and paste Bunny into your > > signature to help him gain world domination! > > ___ > hpx-users mailing list > hpx-users@stellar.cct.lsu.edu > https://mail.cct.lsu.edu/mailman/listinfo/hpx-users > -- == Andreas Schäfer HPC and Supercomputing Institute for Multiscale Simulation Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-20866 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] updated my work so far
Dear Kuo Lu, On 03:03 Thu 23 Jun , Kuo Lu wrote: > I have uploaded my work > <https://github.com/kuolu/LibGeoDecompCommandlineTool> so far. I mainly > studied your bouncingspheres example. And I’m treating the object as a point > in one dimension. > > A few problems for you: > > 1.) I put my folder (including a CMakeList.txt just like every other > examples) in /libgeodecomp/src/examples/ and added > "add_subdirectory(commandlineTools)” in > /libgeodecomp/src/examples/CMakeLists.txt but it still gives me please don't do this. I had sent you a link to a repository[1] which details how to use CMake to build a project which is using LibGeoDecomp (instead of nesting it in LibGeoDecomp's source tree). > CMake Error at src/examples/commandlineTools/CMakeLists.txt:5 > (add_executable): > add_executable called with incorrect number of arguments My best guess is that the variable SOURCES is not set. This is done automatically in LibGeoDecomp's CMake scripts. You should not use such facilities. Instead, just manually list your source files here (i.e. main.cpp). > CMake Error at src/examples/commandlineTools/CMakeLists.txt:6 > (set_target_properties): > set_target_properties Can not find target to add properties to: > libgeodecomp_examples_commandlineTools This fails because line 5 wasn't able to create the target. > CMake Error at src/examples/commandlineTools/CMakeLists.txt:7 > (target_link_libraries): > Cannot specify link libraries for target > "libgeodecomp_examples_commandlineTools" which is not built by this > project. Dito. > Do I need to add more lines in CMakeLists files? If so, which one? Please do not nest your code in LibGeoDecomp's tree. Instead, use a separate directory and set your build up to use LibGeoDecomp as explained in the repository[2]. > 2.) Looking at your > /libgeodecomp/src/examples/bouncingspheres/main.cc, I figured that I > will use SerialSimulator, which means I will need a > Container, initializer, and a writer class just like you did. I > figured this out as I was writing this email while reading the API > documentation… thanks! Yes, the SerialSimulator is a good start. For a start, you should be able to use the BoxCell directly as the container, of a multi-container (see DECLARE_MULTI_CONTAINER_CELL()) once you use different types of bodies. > 3.) what debugger do you use? I cannot use g++ to compile because it > cannot find but cmake keeps giving me > errors. The debugger is largely independent from the compiler. Right now you're using clang++, which is fine. I mostly debug with gdb, but right now your code isn't even compiling so you don't yet need a debugger. > 4.) To be honest, looking at how much I have done so far, I think > I’m far from what I have promised. I thought I know a lot but I > realize that I do not. I want to be honest to myself and deserve a > bad comment in the midterm evaluation, I feel sincerely sorry both > to you and to me. Like I said: making steady progress is now most important. You can still pass the evaluation if you move swiftly. Best -Andreas [1] https://github.com/gentryx/libgeodecomp_website_examples/blob/master/building_cmake_projects/CMakeLists.txt [2] https://github.com/gentryx/libgeodecomp_website_examples/tree/master/building_cmake_projects -- == Andreas Schäfer HPC and Supercomputing Institute for Multiscale Simulation Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-20866 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
[hpx-users] [kuolu1...@gmail.com: updated my work so far]
for reference - Forwarded message from Kuo Lu <kuolu1...@gmail.com> - Date: Thu, 23 Jun 2016 03:03:55 -0400 From: Kuo Lu <kuolu1...@gmail.com> To: Andreas Schäfer <gent...@gmx.de> Subject: updated my work so far X-Mailer: Apple Mail (2.3124) Hi Mr. Schafer, I have uploaded my work <https://github.com/kuolu/LibGeoDecompCommandlineTool> so far. I mainly studied your bouncingspheres example. And I’m treating the object as a point in one dimension. A few problems for you: 1.) I put my folder (including a CMakeList.txt just like every other examples) in /libgeodecomp/src/examples/ and added "add_subdirectory(commandlineTools)” in /libgeodecomp/src/examples/CMakeLists.txt but it still gives me CMake Error at src/examples/commandlineTools/CMakeLists.txt:5 (add_executable): add_executable called with incorrect number of arguments CMake Error at src/examples/commandlineTools/CMakeLists.txt:6 (set_target_properties): set_target_properties Can not find target to add properties to: libgeodecomp_examples_commandlineTools CMake Error at src/examples/commandlineTools/CMakeLists.txt:7 (target_link_libraries): Cannot specify link libraries for target "libgeodecomp_examples_commandlineTools" which is not built by this project. Do I need to add more lines in CMakeLists files? If so, which one? 2.) Looking at your /libgeodecomp/src/examples/bouncingspheres/main.cc, I figured that I will use SerialSimulator, which means I will need a Container, initializer, and a writer class just like you did. I figured this out as I was writing this email while reading the API documentation… thanks! 3.) what debugger do you use? I cannot use g++ to compile because it cannot find but cmake keeps giving me errors. 4.) To be honest, looking at how much I have done so far, I think I’m far from what I have promised. I thought I know a lot but I realize that I do not. I want to be honest to myself and deserve a bad comment in the midterm evaluation, I feel sincerely sorry both to you and to me. Blessings, Kuo - End forwarded message - -- ========== Andreas Schäfer HPC and Supercomputing Institute for Multiscale Simulation Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-20866 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] updated project timeline
Dear Kuo Lu, Thanks for the video call we had on Saturday. I noticed that you haven't yet pushed commits to the repository you created[1]. Are there any questions I can answer for you WRT the parser you'll be writing first? Thanks! -Andreas [1] https://github.com/kuolu/LibGeoDecompCommandlineTool On 10:29 Thu 09 Jun , Kuo Lu wrote: > Hi Mr. Schafer, > > Here is the update: since last time I decided to parse .obj files, I studied > this tutorial > <http://www.opengl-tutorial.org/beginners-tutorials/tutorial-7-model-loading/> > to understand it more (I had some compiling errors as I was following the > tutorial); I then was trying to build a command-line tool that can parse .obj > files in shell script, because I had a conversation with my professor before > about how to build command-line tools and he said “ I never built one, so I > don’t know. Maybe shell script?” So I spent some time studying shell. But > then I realized that integrating shell and c++ is quite hard and it’s not the > project I was picturing, so I went back to study your examples again. I am > picturing to build a environment like matlab or octave, and I just have a > hard time figuring out what language should I use for command-line tools and > how to integrate it with LibGeoDecomp. > > I’m now thinking about either build directly in c++, or use python, > apparently the best language for command-line tools. I’ll do more research > and we can talk about it further in our video call. > > Thank you > > Kuo > > > > > On Jun 9, 2016, at 6:51 AM, Andreas Schäfer <gent...@gmx.de> wrote: > > > > Dear Kuo Lu, > > > > I'm sure you're busy with writing code and figuring out how to best > > parse Blender files. Could you still give me a quick update on what > > you're currently working on? Is there anything I can help with? > > > > Also, did you create a GitHub account and for fork LibGeoDecomp? I > > consider this part of the infrastructure and would like to ask you to > > get this done ASAP so we have a method to integrate your code > > eventually into the trunk. > > > > Thanks! > > -Andreas > > > > > > On 22:54 Wed 01 Jun , Kuo Lu wrote: > >> Hi Mr. Schäfer, > >> > >> Thank you for your patience. I know that I have a slow start and I > >> apologize for that. I have been taking a summer course and it takes almost > >> two thirds of my day everyday. Fortunately my summer class ends by the end > >> of this week so I will be able to catch up my work starting this weekend. > >> And also I adjusted my project timeline > >> <https://docs.google.com/document/d/1mDPVtyarMGn48kkHCPns7ayrMn4wOzRTgIEsRiFz6WU/edit?usp=sharing> > >> a little bit based on my current situation. > >> > >> Thank you and best regards, > >> > >> Kuo > > > > -- > > == > > Andreas Schäfer > > HPC and Supercomputing > > Institute for Multiscale Simulation > > Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany > > +49 9131 85-20866 > > PGP/GPG key via keyserver > > http://www.libgeodecomp.org > > == > > > > (\___/) > > (+'.'+) > > (")_(") > > This is Bunny. Copy and paste Bunny into your > > signature to help him gain world domination! > -- == Andreas Schäfer HPC and Supercomputing Institute for Multiscale Simulation Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-20866 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] Just a student who wants to get involved in a GSoC project
Dear Cristian, we always appreciate contributors so you're most welcome to aid with the development of any of our projects (e.g. HPX and LibGeoDecomp). independently from GSoC. Kind regards -Andreas On 02:45 Fri 25 Mar , Cristian Dragu wrote: > Hello Andreas, > > Thank you for your reply. I will have a look over the archives you gave me. > Considering I'm using my actual time learning CUDA C I will decide which > one of the two projects fits better with my knowledge. One important > questions is: Would I be able to continue working on this project by using > the archives, any other communication channels (IRC, mail list, and > others), even if I am not sending a proposal? (This means that I will not > participate to GSoC). I say this only because the deadline is today, I > didn't work on this project until now, and I have 2 homework to finish in 3 > days for faculty. I am eager to learn new things about parallel and > distributed algorithms, parallel calculations and related stuff, and I > would be grateful if I could continue to work by my schedule (faculty has a > great effect on my schedule) on one of the two projects. > > If you have some advices, they are welcomed. > > Best regards, > Cristian Dragu > > == > > (\__/) > (+'.'+) > (")_(") > > I will copy this bunny because it's interesting. > > On Fri, Mar 25, 2016 at 12:26 AM, Andreas Schäfer <gent...@gmx.de> wrote: > > > Cristian, > > > > thanks for reaching out to me directly. I'm CC'ing hpx-users here and > > I'd suggest you to subscribe to that list yourself as we keep all of > > our GSoC discussions on this list. > > > > On 11:32 Thu 24 Mar , Cristian Dragu wrote: > > > My name is Cristian Dragu. I am a student at Faculty of Automatic Control > > > and Computers, University Politehnica of Bucharest. > > > > > > I would like to get involved in one of the projects you are mentoring at > > > GSoC. The two projects I am interested in are Implement a Map/Reduce > > > Framework and Newtonian physics sandbox. > > > > > > During the past semester I have studied the Parallel and distributed > > > algorithms (completed some homework using MPI, OpenMP and > > ReplicatedWorkers > > > (Map/Reduce) with threading in Java). > > > > > > I would like to talk to you about these two projects, because I have to > > > choose one of them, but first I must know more things about them. > > > > > > Firstly, I have looked over Hadoop since it's a framework somehow related > > > to the Map/Reduce Framework. I would like you to tell me how different, > > > less/more complicated the framework should be compared to Hadoop. > > > I have started reading some tutorials about Hadoop and looked over his > > > arhitecture: > > > > > http://blog.raremile.com/wp-content/uploads/2012/09/hadoop_architecture.jpg > > > > We've covered this project before, so you might want to check the > > mailing list's archives[1] to get more info. I don't think it would be > > reasonable to expect anything on par with Hadoop (which is production > > quality software) from a short GSoC project. > > > > > Secondly, I would like you to tell me more things about Newtonian physics > > > sandbox. It looks like a nice project, but I have exactly no idea how to > > > start it, because probably I misunderstood the project's explanation. The > > > only thing I do at this moment is to watch some tutorial for CUDA > > > (LibGeoDecomp supports CUDA, so I thought that if I learn some CUDA C it > > > might come in help). > > > > For speed: please take a look at my exchange with Aleksandr Sofienko > > on this project on the hpx-users list (archive for March). Please feel > > free if you have any further questions. > > > > Kind regards > > -Andreas > > > > [1] http://mail.cct.lsu.edu/pipermail/hpx-users/ > > > > > > -- > > == > > Andreas Schäfer > > HPC and Grid Computing > > Department of Computer Science 3 > > Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany > > +49 9131 85-27910 > > PGP/GPG key via keyserver > > http://www.libgeodecomp.org > > == > > > > (\___/) > > (+'.'+) > > (")_(") > > This is Bunny. Copy and paste Bunny into your > > signature to help him gain world domination! > > -- == Andreas Schäfer HPC and Grid Computing Department of Computer Science 3 Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-27910 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] Just a student who wants to get involved in a GSoC project
Cristian, thanks for reaching out to me directly. I'm CC'ing hpx-users here and I'd suggest you to subscribe to that list yourself as we keep all of our GSoC discussions on this list. On 11:32 Thu 24 Mar , Cristian Dragu wrote: > My name is Cristian Dragu. I am a student at Faculty of Automatic Control > and Computers, University Politehnica of Bucharest. > > I would like to get involved in one of the projects you are mentoring at > GSoC. The two projects I am interested in are Implement a Map/Reduce > Framework and Newtonian physics sandbox. > > During the past semester I have studied the Parallel and distributed > algorithms (completed some homework using MPI, OpenMP and ReplicatedWorkers > (Map/Reduce) with threading in Java). > > I would like to talk to you about these two projects, because I have to > choose one of them, but first I must know more things about them. > > Firstly, I have looked over Hadoop since it's a framework somehow related > to the Map/Reduce Framework. I would like you to tell me how different, > less/more complicated the framework should be compared to Hadoop. > I have started reading some tutorials about Hadoop and looked over his > arhitecture: > http://blog.raremile.com/wp-content/uploads/2012/09/hadoop_architecture.jpg We've covered this project before, so you might want to check the mailing list's archives[1] to get more info. I don't think it would be reasonable to expect anything on par with Hadoop (which is production quality software) from a short GSoC project. > Secondly, I would like you to tell me more things about Newtonian physics > sandbox. It looks like a nice project, but I have exactly no idea how to > start it, because probably I misunderstood the project's explanation. The > only thing I do at this moment is to watch some tutorial for CUDA > (LibGeoDecomp supports CUDA, so I thought that if I learn some CUDA C it > might come in help). For speed: please take a look at my exchange with Aleksandr Sofienko on this project on the hpx-users list (archive for March). Please feel free if you have any further questions. Kind regards -Andreas [1] http://mail.cct.lsu.edu/pipermail/hpx-users/ -- == Andreas Schäfer HPC and Grid Computing Department of Computer Science 3 Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-27910 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] Interest in 'Implement a plugin mechanism for thread scheduling' and 'create wrapper for Region class of LibGeoDecomp' for GSOC 2016
Hey Satyaki, On 00:07 Thu 10 Mar , Satyaki Upadhyay wrote: > *Create a wrapper for Region class of LibGeoDecomp to be used with BGL* > I loved the idea of this project because it involves both C++ (the language > I'm most comfortable with) and graph theory (which I've been using regulary > for the past two years). As a consequence of my regular participation in > algorithmic programming competitions, I am reasonably familiar with graph > theory and its algorithms (eg BFS/DFS/Dijkstra/Floyd-WarshallTarjan's to > name a few). I would like to know how to get started with this and would > like to discuss what needs to be done in some more detail. I'm happy to hear that this project idea sparked your interest. LibGeoDecomp's Region class could be used to store the list of vertices as well as the edge set in BGL. Depending on the scenario (properties of the graph and operations performed on it) this might be efficient, or not. Right now BGL relies on standard containers (std::vector in adjacency_matrix, std::list in adjacency_list) for these tasks. Some background info on LibGeoDecomp's Region: as our ideas page says it's a fast runlength-compressed set for 1D/2D/3D coordinates. It's a) very fast for iteration, b) logical operations (e.g. intersection, union), and c) has a low memory footprint. My suggestions to get started with this idea would be: - take a look at the BGL source, starting with examples and then dive into the library source. Your focus should be: which operations are performed on the standard containers? In which areas of the code are they used? - take a look at libgeodecomp/src/geometry/region.h and, more importantly, libgeodecomp/src/geometry/test/unit/region.h. The tests should give you an idea what Region is capable of and how it's used. libgeodecomp/src/geometry/test/unit/regionbasedadjacency.h and libgeodecomp/src/geometry/regionbasedadjacency.h should give you an idea how Region<2> can store an adjacency list. - finally: how can the operations performed on std::list/vector in BGL be translated into operations on Region? make a brief list of the required operations and their approxiamate counterparts in Region. Please feel free to ask further questions. Cheers -Andreas -- ========== Andreas Schäfer HPC and Grid Computing Department of Computer Science 3 Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-27910 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] Program failing when not compiling with -O3
On 22:23 Tue 17 Nov , Thomas Heller wrote: > Am 17.11.2015 10:10 nachm. schrieb "Hartmut Kaiser" < > hartmut.kai...@gmail.com>: > > > > > > > So, Zach and I have hit the same error when interfacing dg-swem with > > > HPX+LGD, but here we weren't able to work around it via adding -O3. > > > Any ideas for another workaround? > > > > > > Thomas said he's suspecting a bug in HPX, but I can't really figure > > > out what's going wrong. > > > > As said, I don't understand what you're doing with that code. The > > component-registry stuff is very subtle to get right and I don't see any > > obvious reason you create your own. > > Does it work when you use the HPX provided registration macros? Yes. I cobbled together a macro which in turn instantiates the corresponding HPX macros. This works and and might be a makeshift solution for our current DGSWEM branch. I'd like to avoid using this for a release though as I'd really like to avoid forcing the user to instantiate macros just for the HPX backend. Cheers -Andi -- == Andreas Schäfer HPC and Grid Computing Department of Computer Science 3 Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-27910 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
Re: [hpx-users] Program failing when not compiling with -O3
So, Zach and I have hit the same error when interfacing dg-swem with HPX+LGD, but here we weren't able to work around it via adding -O3. Any ideas for another workaround? Thomas said he's suspecting a bug in HPX, but I can't really figure out what's going wrong. Thanks! -Andreas On 23:49 Sun 15 Nov , Andreas Schäfer wrote: > Heya, > > I've recently seen some of my HPX applications terminating right > before hpx_init() is called. Strangely the build system did seem to > determine whether the same code would either run flawlessly, or crash. > > Please find attached a (sort of) minimal example that reproduces the > error. Below is a log of me compiling the code manually. The only > difference here is that for the first run I did add -O3, for the > second run that option was omitted. Any ideas what's going on here? > > > gentryx@neuromancer ~ $ time (rm -f test_hpx && > > /usr/lib64/ccache/bin/g++-4.9.3 -O3-L/home/gentryx/local_install/lib > > -lhpx test_hpx.cpp -o test_hpx -std=c++11 > > -I/home/gentryx/local_install/include > > -I/home/gentryx/local_install/include/hpx/external -lboost_chrono > > -lboost_date_time -lboost_filesystem -lboost_program_options -lboost_system > > -lboost_thread -lhpx_init -lhpx -ldl -lrt && > > LD_LIBRARY_PATH=/home/gentryx/local_install/lib && > > LD_LIBRARY_PATH=/home/gentryx/local_install/lib ./test_hpx) > > ok > > > > real0m28.162s > > user0m27.490s > > sys 0m0.630s > > > > 23:43:58 - 0 > > gentryx@neuromancer ~ $ time (rm -f test_hpx && > > /usr/lib64/ccache/bin/g++-4.9.3-L/home/gentryx/local_install/lib -lhpx > > test_hpx.cpp -o test_hpx -std=c++11 -I/home/gentryx/local_install/include > > -I/home/gentryx/local_install/include/hpx/external -lboost_chrono > > -lboost_date_time -lboost_filesystem -lboost_program_options -lboost_system > > -lboost_thread -lhpx_init -lhpx -ldl -lrt && > > LD_LIBRARY_PATH=/home/gentryx/local_install/lib && > > LD_LIBRARY_PATH=/home/gentryx/local_install/lib ./test_hpx) > > terminate called after throwing an instance of > > 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector > > >' > > what(): attempt to insert a GVA with an invalid type, > > gid({00010001, 0001}), gva(({0001, > > } component_invalid[-1] 1 0xa41640 0)), > > locality({0001, }): HPX(bad_parameter) > > > > real0m22.700s > > user0m21.910s > > sys 0m0.740s > > > > 23:44:41 - 134 > > Thanks! > -Andreas > > > -- > == > Andreas Schäfer > HPC and Grid Computing > Department of Computer Science 3 > Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany > +49 9131 85-27910 > PGP/GPG key via keyserver > http://www.libgeodecomp.org > == > > (\___/) > (+'.'+) > (")_(") > This is Bunny. Copy and paste Bunny into your > signature to help him gain world domination! > #include > #include > > #include > #include > #include > #include > > namespace { > template > class hpx_plugin_exporter_factory; > > template > class init_registry_factory_static; > > template > class hpx_plugin_exporter_registry; > > } > > namespace LibGeoDecomp { > > /** > * Instantiate this template to ensure the instantiation of an HPX > * component template is actually registered. See HPXReceiver for an > * example on how to use this class and its assorted macros. > */ > template > class HPXComponentRegistrator > { > public: > virtual ~HPXComponentRegistrator() > {} > > static > hpx::components::component_factory<hpx::components::simple_component > > *instanceA; > static hpx::components::component_registry< > hpx::components::simple_component, > ::hpx::components::factory_check> *instanceB; > > virtual > hpx::components::component_factory<hpx::components::simple_component > > *foo1() > { > instanceA = new > hpx::components::component_factory<hpx::components::simple_component > >(0, 0, false); > return instanceA; > } > > virtual hpx::components::component_registry< > hpx::components::simple_component, > ::hpx::components::factory_check> *foo2() > { > instanceB = new hpx::components::component_registry< > hpx::components::simple_component, > ::hpx::compone
Re: [hpx-users] Program failing when not compiling with -O3
On 15:08 Tue 17 Nov , Hartmut Kaiser wrote: > Why do you need your own component registry etc.? I can't use macros because a user would need to instantiate them manually. Users forget about that stuff, so we do it for them. Cheers -Andi -- == Andreas Schäfer HPC and Grid Computing Department of Computer Science 3 Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany +49 9131 85-27910 PGP/GPG key via keyserver http://www.libgeodecomp.org == (\___/) (+'.'+) (")_(") This is Bunny. Copy and paste Bunny into your signature to help him gain world domination! signature.asc Description: Digital signature ___ hpx-users mailing list hpx-users@stellar.cct.lsu.edu https://mail.cct.lsu.edu/mailman/listinfo/hpx-users