Re: MEMD (SCAFACOS) interfacing issue

[Jean-Noël Grad]

Hi Kheerti and Florian,

Just a quick note to let you know that an interface to MEMD is now 
available in my fork of ScaFaCoS. You can find more details here:
https://github.com/espressomd/espresso/issues/4536#issuecomment-1190635035

I know this new development arrives quite late in this thread. If anyone 
is still working on this and is willing to invest developer time, feel 
free to open a MEMD tracking ticket on the ScaFaCoS project, so that 
other users can join and help out. While I won't be able to donate more 
developer time to this feature, there was another user interested in the 
method on GitHub. It's unclear at the moment, for which range of input 
parameters the MEMD method works, and it is prone to memory errors and 
integer overflow errors.

Best regards,
Jean-Noël Grad

On 2021-09-25 22:23, Florian Fahrenberger wrote:
> Hi Kheerti,
>
> I was the PhD student at the time who implemented the MEMD method in 
> the
> ScaFaCoS library. However, I am not in scientific research anymore and
> don't think I will be easily able to debug the code, at least not on an
> acceptable timeline.
>
> May I ask why you picked the MEMD algorithm out of all the available?
> Because it is a very special algorithm that does provide some benefits
> in very special and particular cases, mostly a smoothly varying
> inhomogenous dielectric background constant. But this feature is not
> available in the ScaFaCoS library, since you don't have an interface to
> set the local dielectric background anyways.
>
> So honestly, even though I implemented the code and got to like the
> algorithm with time, I would probably recommend you to pick a different
> algorithm for your Coulomb solver. Most other methods are much more
> established, at least as fast, and are more regularly updated. Last I
> looked, the P3M method is the most commonly used for coarse grained 
> soft
> matter physics applications, and it is the one I would probably use
> myself if I plugged the library into any of my codes.
>
> If you desperately need the MEMD method to work, I can take another 
> look
> at it, but it has been years since I even compiled this, and I cannot
> guarantee that this would be fast.
>
> I hope this answer helps you to some degree and wish you the best of
> luck in your research,
>
> Florian
>
>
> Am 25.09.2021 um 16:13 schrieb Keerthi Radhakrishnan 150503 via 
> ESPResSo
> developers' discussions:
> > Dear all,
> >
> > I am a PhD student currently pursuing research in soft matter physics.
> > My work mostly deals with charged polymeric systems. Presently, my
> > project at hand demands the incorporation of dielectric variation 
> > within the solvent in the  presence of explicit charges. For that I
> > resort to the MEMD solver included in SCAFACOS library for the
> > electrostatics. This is the code snippet from my pypresso script where
> > it incubates MEMD following  the instructions given in the espresso 
> > manual.
> >
> > from espressomd.electrostatics import Scafacos
> > from espressomd.electrostatics import Scafacos
> > scafacos = Scafacos(prefactor=1, method_name="memd",
> > method_params={"timestep": 0.005, "mesh": 1, "temperature":1,
> > "bjerrum_length":3})
> > system.actors.add(scafacos)
> >  
> > Following this I get an  error regarding the unrecognition of the 
> > passed
> > argument, which I have carefully taken the argument names mentioned in
> > SCAFACOS manual itself for MEMD Solver-specific parameters. If this is
> > not the case of how arguments are named, then where in espresso
> > documentations is anything regarding it given.
> >
> >
> > The ERROR is follows:
> > New handle created!
> > terminate called after throwing an instance of 'std::runtime_error'
> >   what():  interface (parser): error in parameter string at 
> > 'timestep'!
> > [Langevin:04113] *** Process received signal ***
> > [Langevin:04113] Signal: Aborted (6)
> > [Langevin:04113] Signal code:  (-6)
> > [Langevin:04113] [ 0]
> > /lib/x86_64-linux-gnu/libc.so.6(+0x3efd0)[0x7f58ac287fd0]
> > [Langevin:04113] [ 1]
> > /lib/x86_64-linux-gnu/libc.so.6(gsignal+0xc7)[0x7f58ac287f47]
> > [Langevin:04113] [ 2]
> > /lib/x86_64-linux-gnu/libc.so.6(abort+0x141)[0x7f58ac2898b1]
> > [Langevin:04113] [ 3]
> > /usr/lib/x86_64-linux-gnu/libstdc++.so.6(+0x8c957)[0x7f58a9d42957]
> > [Langevin:04113] [ 4]
> > /usr/lib/x86_64-linux-gnu/libstdc++.so.6(+0x92ae6)[0x7f58a9d48ae6]
> > [Langevin:04113] [ 5]
> > /usr/lib/x86_64-linux-gnu/libstdc++.so.6(+0x92b21)[0x7f58a9d48b21]
> > [Langevin:04113] [ 6]
> > /usr/lib/x86_64-linux-gnu/libstdc++.so.6(+0x92d54)[0x7f58a9d48d54]
> > [Langevin:04113] [ 7]
> > /home/system1/espresso/test_build2/src/scafacos/Scafacos.so(_ZN8Scafacos8Scafacos16parse_parametersERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE+0x68)[0x7f58a1931938]
> > [Langevin:04113] [ 8]
> > /home/system1/espresso/test_build2/src/scafacos/Scafacos.so(_ZN8Scafacos8ScafacosC1ERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEEP19ompi_communicator_tS8_+0xca)[0x7f58a1931a1a]
> > [Langevin:04113] [ 9]
> > /home/system1/espresso/test_build2/src/core/EspressoCore.so(+0x25261d)[0x7f58aa79d61d]
> > [Langevin:04113] [10]
> > /home/system1/espresso/test_build2/src/core/EspressoCore.so(_ZN8Scafacos14set_parametersERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEES7_b+0x3a6)[0x7f58aa79dab6]
> > [Langevin:04113] [11]
> > /home/system1/espresso/test_build2/src/python/espressomd/scafacos.so(+0xb651)[0x7f58084c3651]
> > [Langevin:04113] [12]
> > /home/system1/espresso/test_build2/src/python/espressomd/electrostatics.so(+0x39c3d)[0x7f5808704c3d]
> > [Langevin:04113] [13]
> > /home/system1/espresso/test_build2/src/python/espressomd/electrostatics.so(+0x39e0d)[0x7f5808704e0d]
> > [Langevin:04113] [14]
> > /home/system1/espresso/test_build2/src/python/espressomd/electrostatics.so(+0x4ff46)[0x7f580871af46]
> > [Langevin:04113] [15]
> > /home/system1/espresso/test_build2/src/python/espressomd/actors.so(+0xd47d)[0x7f585c92947d]
> > [Langevin:04113] [16]
> > /home/system1/espresso/test_build2/src/python/espressomd/actors.so(+0xd64d)[0x7f585c92964d]
> > [Langevin:04113] [17]
> > /home/system1/espresso/test_build2/src/python/espressomd/actors.so(+0x19b06)[0x7f585c935b06]
> > [Langevin:04113] [18]
> > /home/system1/espresso/test_build2/src/python/espressomd/script_interface.so(+0x135ec)[0x7f589df195ec]
> > [Langevin:04113] [19]
> > /home/system1/espresso/test_build2/src/python/espressomd/actors.so(+0xd47d)[0x7f585c92947d]
> > [Langevin:04113] [20]
> > /home/system1/espresso/test_build2/src/python/espressomd/actors.so(+0xd64d)[0x7f585c92964d]
> > [Langevin:04113] [21]
> > /home/system1/espresso/test_build2/src/python/espressomd/actors.so(+0x1146a)[0x7f585c92d46a]
> > [Langevin:04113] [22]
> > /usr/bin/python3(_PyObject_FastCallKeywords+0x19c)[0x5a9d2c]
> > [Langevin:04113] [23] /usr/bin/python3[0x50a533]
> > [Langevin:04113] [24]
> > /usr/bin/python3(_PyEval_EvalFrameDefault+0x444)[0x50bf44]
> > [Langevin:04113] [25] /usr/bin/python3[0x507cd4]
> > [Langevin:04113] [26] /usr/bin/python3(PyEval_EvalCode+0x23)[0x50ae13]
> > [Langevin:04113] [27] /usr/bin/python3[0x635262]
> > [Langevin:04113] [28] 
> > /usr/bin/python3(PyRun_FileExFlags+0x97)[0x635317]
> > [Langevin:04113] [29]
> > /usr/bin/python3(PyRun_SimpleFileExFlags+0x17f)[0x638acf]
> > [Langevin:04113] *** End of error message ***
> > Aborted (core dumped)
> >
> >
> > Following, this I also figured that the interface file in scafacos: 
> >   scafacos-fcs-0.1/src/FCSInterface.c  - line 1633 the part where
> > arguments for memd is defined is commented. Don't know if this has
> > anything to do with the existing issue. 
> > An insight into the root of this problem and a practical resolve to it
> > would be highly appreciated.
> >  
> > Regards
> > Keerthi