Articulated rigid body dynamics

[tkoziara]

As transpired in #7 it would be useful to include into Parmec implicit rigid links to allow modelling of articulated chains of bodies. This can be done via the same mechanism as in Solfec (general but slow) or by means of a recursive formulation (fast). This issue will explore and lead to a practical implementation of this functionality. Mentioning @bencannell and @stevebrasier who prompted this development.

[tkoziara]

@bencannell @stevebrasier I am thinking of starting with something as simple as the BALL_JOINT command specified below. I will be able to copy some of the Solfec functionality to implement this. It should serve your needs. Do feel free to provide comments/suggestions, while I start looking into this.

jnum = BALL_JOINT (part1, point | part2)
• jnum - joint number
• part1 - first particle number
• point - coincident point (x,y,z)
• part2 - optional second particle number; default: not specified (in this case part1 is fixed in space at point)

Create a ball joint between two particles at a coincident point. The joint is modeled as an algebraic constraint, without affecting the stable time step. Any number of joints between any number of particles can be added. A linear system is solved at every time step to calculate joint forces.

[bencannell]

Hi @tkoziara,

Yes this looks appropriate, assuming this can be coupled with the TORSION_SPRING functionality?

[tkoziara]

Yes: joints and springs will be independent functionalities - able to be coupled without restrictions. I am looking into including a suitable linear solver package into parmec now, to serve the related algebra needs. I will start on this following that.

[bencannell]

OK, great, thank you for the update :)

Sorry I haven't tested the current TORSION_SPRING functionality yet. I'll do so this week.

[tkoziara]

Take your time; I will be progressing other things in the meantime:)

[tkoziara]

@bencannell @stevebrasier , fyi, this is now functional: http://parmes.org/blog/parmec_ball_joint.html

[bencannell]

Hi @tkoziara

We have begun testing this. I'm getting the following error when compiling changeset 100711a.

g++ -O2 -Wno-narrowing -DISPC_USE_OMP tasksys.cpp -c -o objs4/tasksys.o
g++ -DREAL=4 -Iobjs4 -Iobjs4 -O2 -Wno-narrowing mem.cpp -c -o objs4/mem.o
g++ -DREAL=4 -Iobjs4 -Iobjs4 -O2 -Wno-narrowing map.cpp -c -o objs4/map.o
g++ -DREAL=4 -Iobjs4 -Iobjs4 -O2 -Wno-narrowing mesh.cpp -c -o objs4/mesh.o
g++ -DREAL=4 -Iobjs4 -Iobjs4 -O2 -Wno-narrowing timeseries.cpp -c -o objs4/timeseries.o
g++ -DREAL=4 -Iobjs4 -O2 -Wno-narrowing -I. -std=c++11 joints.cpp -c -o objs4/joints.o
cc1plus: error: unrecognized command line option "-std=c++11"
cc1plus: warning: unrecognized command line option "-Wno-narrowing"
make: *** [objs4/joints.o] Error 1

Do you have a similar error?

Thanks,

Ben

[tkoziara]

Hi Ben. Which version of g++ are you using? It does not seem to support C++11, which is needed to compile the joints functionality (a linear solver within it uses it). Tomek

[bencannell]

g++ (GCC) 4.4.7 20120313 (Red Hat 4.4.7-16)
Copyright (C) 2010 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

[bencannell]

Which version would you suggest we use?

[tkoziara]

Perhaps try something a little bit higher. See:
http://www.gnu.org/software/gcc/projects/cxx-status.html
I am using 7.2.0 at the moment.

[bencannell]

Hi @tkoziara,

We've now got a more appropriate version of g++ installed and can successfully compile the updated parmec source code.

I notice that the parmec executable will only run on a single core. Do you observe the same behaviour?

Many thanks,

Ben

[tkoziara]

Not on my laptop - just checked. What's the symptom for you? How are you testing this?