Re: [ESPResSo-users] Fwd: Brownian dynamics

[David Schwörer]

Hi Dusan,

I have fixed the first two points, the code is on github - thanks for
reporting the bugs.

Concerning point 3, the BD feature does not use the GPU, it always runs
only on the CPU. The SD on the other hand, only runs on GPUs, a CPU
implementation is not available.
If you want to compute the forces on the GPU then I am the wrong person
to ask.

The BD could not be ported to the GPU, as the computation is very
simple, and the transfer to and from the device are slower than the
computation on the CPU.
The SD code could be ported to the CPU, but that would require a
significant amount of work ...

Kind Regards,
David

On 18/04/16 16:17, Dudo wrote:
> Hi David,
> 
> thanks a lot. It compiles and runs the simple script.
> I'm able also to run my earlier scripts with 'integrate_sd'.
> I have a few comments:
> 1) 'require_feature BD' says that BD is not compiled and aborts the
> script, hence avoid using..
> 2) 'puts [thermostat]' exists with error
> 3) how much faster is it with GPU for bonded systems (did anybody try)?
> 
> have a nice day :-)
> Dusan
> 
> 
> 
> 
> On Fri, Apr 8, 2016 at 4:23 PM, Lei Liu <address@hidden
> <mailto:address@hidden>> wrote:
> 
>     Hello David,
> 
>     Yes, I have ready succeeded to run the simple script on my pc.
>     I would like to do a little bit more tests on this integrator, as
>     well as its thermostat. 
> 
>     Thanks a LOT for your help.
> 
>     With my best wishes
>     Lei
> 
> 
> 
>     On Fri, Apr 8, 2016 at 10:38 PM, David Schwörer
>     <address@hidden <mailto:address@hidden>>
>     wrote:
> 
>         Hi all,
> 
>         I changed the ifdefs, now it should be able compile with #define
>         BD, but
>         without SD/CUDA/ARPACK. See:
>         https://github.com/espressomd/espresso/pull/579
> 
>         I haven't used the build system, but with:
>         ./bootstrap.sh
>         ./configure --with-python-interface=no && make
>         ./Espresso bd_simple.tcl
> 
>         I was able to run the bd integrator.
> 
>         If the python feature is needed, I could have a look at it ...
>         I am not sure about cmake - it failed at TCLTK - which I have
>         not installed.
> 
>         Kind Regards,
>         David
> 
>         On 08/04/16 08:45, Lei Liu wrote:
>         > Hi Peter,
>         >
>         > thanks for your info.
>         > If it turns out that it not easy to compile it with CPU, that could 
> be
>         > my solution.
>         >
>         > Best regards
>         > Lei
>         >
>         > On Fri, Apr 8, 2016 at 4:38 PM, Peter Košovan
>         > <address@hidden <mailto:address@hidden>
>         <mailto:address@hidden
>         <mailto:address@hidden>>> wrote:
>         >
>         >     Hi all,
>         >
>         >     I just want to add that in the last few days with my student 
> Jakub
>         >     we were also trying to compile espresso with SD and encountered
>         >     similar problems. But we managed to compile it with cuda and 
> arpack,
>         >     so the origin of this problem is probably different.
>         >
>         >     Could perhaps David comment on that?
>         >
>         >     With best regards,
>         >
>         >     peter
>         >
>         >     On Fri, Apr 8, 2016 at 9:27 AM, Lei Liu <address@hidden 
> <mailto:address@hidden>
>         >     <mailto:address@hidden <mailto:address@hidden>>> wrote:
>         >
>         >
>         >         ---------- Forwarded message ----------
>         >         From: *Lei Liu* <address@hidden <mailto:address@hidden>
>         <mailto:address@hidden <mailto:address@hidden>>>
>         >         Date: Fri, Apr 8, 2016 at 4:23 PM
>         >         Subject: Re: [ESPResSo-users] Brownian dynamics
>         >         To: David Schwörer <address@hidden
>         <mailto:address@hidden>
>         >         <mailto:address@hidden
>         <mailto:address@hidden>>>
>         >
>         >
>         >         Dear all,
>         >
>         >         I am also interested in the (over-damped) Brownian
>         dynamics.
>         >
>         >         Based on David's suggestion, I downloaded current
>         developing
>         >         version of Espresso,
>         >         and want to compile it with SD integrator.
>         >
>         >         Since I don't have experiences with cuda,
>         >         I commented the requirements of cuda and arpack for SD
>         in  the
>         >         file "/src/features.def",
>         >         as well as the error message output at the head of the
>         file
>         >         "/src/core/integrate_sd.hpp" ,
>         >         boot-trapped in the source code directory, and then
>         tried to
>         >         configure and make it.
>         >
>         >         It went through the compiling in the core directory,
>         >         but reported following errors in the file
>         "integrate_sd_tcl.cpp":
>         >
>         >         error: 'sd_set_particles_apart' was not declared in
>         this scope.
>         >         error: 'sd_test' was not declared in this scope
>         >         error: 'integrate_sd' was not declared in this scope.
>         >
>         >         I don't understand what caused the problem here,
>         >         because I do find the statement '#include
>         "integrate_sd.hpp" '
>         >         at the head of file "integrate_sd_tcl.cpp".
>         >
>         >         If anyone known the reason or has some experiences about
>         >         compiling SD integrator with cpu,
>         >         I would appreciate quite a lot.
>         >
>         >         Best regards
>         >         Lei
>         >         ------------------------------
>         >         Korea Institute for Advanced Research
>         >         85 Hoegiro, Dongdaemun-gu
>         >         Seoul 02455
>         >         Republic of Korea
>         >
>         >
>         >
>         >         On Fri, Dec 18, 2015 at 7:27 AM, David Schwörer
>         >         <address@hidden
>         <mailto:address@hidden>
>         >         <mailto:address@hidden
>         <mailto:address@hidden>>> wrote:
>         >
>         >             Hi,
>         >
>         >             Together with the Stokesian Dynamics integrator I
>         >             implemented a BD
>         >             integrator as well.
>         >             It should work, if you compile with the SD
>         enabled, you
>         >             should be able
>         >             to use
>         >
>         >             part 0 pos 0 0 0
>         >             setmd time_step 0.1
>         >             setmd sd_visc 1
>         >             setmd sd_radius 1
>         >             integrate_sd 1
>         >             puts [part 0 print pos]
>         >
>         >             If you cannot meet the requirements for SD (CUDA
>         etc.), I
>         >             could try to
>         >             patch it, that BD can be compiled without SD.
>         >
>         >             Kind Regards,
>         >             David
>         >
>         >             On 12/17/2015 01:20 PM, Dudo wrote:
>         >             > Hi,
>         >             >
>         >             > please, is there a project or a feature of Espresso
>         >             running Brownian
>         >             > dynamics?
>         >             > I figured out, that just setting mass to zero
>         doesn't work :-)
>         >             > In the integrator (integrate.cpp) there is also
>         velocity
>         >             scaling which
>         >             > goes
>         >             > to infinity with the mass zero..
>         >             >
>         >             > Kind regards,
>         >             > Dusan
>         >             >
>         >             > --
>         >             > _____________________
>         >             > Faculté de biologie et de médecine
>         >             > Center for Integrative Genomics
>         >             > GENOPODE UNIL-SORGE
>         >             > Université de Lausanne
>         >             > CH1015 Lausanne, Suisse [+]
>         >             >
>         >             > RG: https://www.researchgate.net/profile/Dusan_Racko
>         >             > Scholar: https://goo.gl/oRVrGV
>         >             >
>         >             > Polymer Institute of the Slovak Academy of Sciences
>         >             > Dubravska cesta 3
>         >             > 845 41 Bratislava, Slovak Republic [‡]
>         >             > tel: +421 2 3229 4321
>         >
>         >
>         >
>         >
>         >
>         >
>         >
>         >
>         >     --
>         >     Dr. Peter Košovan
>         >
>         >     Department of Physical and Macromolecular Chemistry
>         >     Faculty of Science, Charles University in Prague, Czech
>         Republic
>         >
>         >     Katedra fyzikální a makromolekulární chemie
>         >     Přírodovědecká fakulta Univerzity Karlovy v Praze
>         >
>         >     www.natur.cuni.cz/chemistry/fyzchem/
>         <http://www.natur.cuni.cz/chemistry/fyzchem/>
>         >     <http://www.natur.cuni.cz/chemistry/fyzchem/>
>         >     Tel. +420221951290 <tel:%2B420221951290>
>         >     Fax +420224919752 <tel:%2B420224919752>
>         >
>         >
> 
> 
> 
> 
> 
> -- 
> _____________________
> Faculté de biologie et de médecine
> Center for Integrative Genomics
> GENOPODE UNIL-SORGE
> Université de Lausanne
> CH1015 Lausanne, Suisse [+]
> 
> RG: https://www.researchgate.net/profile/Dusan_Racko
> Scholar: https://goo.gl/oRVrGV
> 
> Polymer Institute of the Slovak Academy of Sciences
> Dubravska cesta 3
> 845 41 Bratislava, Slovak Republic [‡]
> tel: +421 2 3229 4321