search

CEED / Laghos

High-order Lagrangian Hydrodynamics Miniapp
http://ceed.exascaleproject.org/miniapps
BSD 2-Clause "Simplified" License
187 stars 60 forks source link

Add Rayleigh-Taylor Instability Problem to rom branch #114

Closed siuwuncheung closed 3 years ago

siuwuncheung commented 4 years ago

The Rayleigh-Taylor Instability Problem can be run as follows: ./laghos -p 7 -m data/rt2D.mesh -tf 4 -rs 1 -ok 4 -ot 3 -vis -pa The figures are taken at t = 4.0, and the formation of the ”mushroom cap” can be seen.

Velocity at t = 4.0 velocity

Energy at t = 4.0 energy

Density at t = 4.0 density

dylan-copeland commented 3 years ago

Are the added mesh files rt2D.mesh, square01_quad_unstr.mesh, square01_tri.mesh necessary? I don't see where they are used.

dylan-copeland commented 3 years ago

@siuwuncheung The regression tests all seem to be failing, and I suppose it is because of small differences (e.g. number of timesteps) due to round-off level changes caused by updating the FOM code from the master branch of Laghos. If so, that is fine, but can you please report whether you have verified this is the case? It is a lot of work to check every regression test output file to verify that the change is acceptable, so I think it is fine if just one person (usually the PR author) does this and reports the results.

chldkdtn commented 3 years ago

@siuwuncheung please see if you get a seg fault when running the following simulation:

./laghos -p 7 -m data/rt2D.mesh -tf 4 -rs 1 -ok 4 -ot 3 -offline -romsvds -ef 0.9999 -nwinsamp 5 -romos -rostype load -romsrhs -writesol -visit

I suspect that it is something to do with the end or beginning of time window exactly coincides with the end of time steps.

siuwuncheung commented 3 years ago

Are the added mesh files rt2D.mesh, square01_quad_unstr.mesh, square01_tri.mesh necessary? I don't see where they are used.

The mesh file rt2D.mesh is used for the 2D Rayleigh-Taylor instability problem. The mesh files square01_quad_unstr.mesh and square01_tri.mesh are not necessary and can be removed.

siuwuncheung commented 3 years ago

@siuwuncheung please see if you get a seg fault when running the following simulation:

./laghos -p 7 -m data/rt2D.mesh -tf 4 -rs 1 -ok 4 -ot 3 -offline -romsvds -ef 0.9999 -nwinsamp 5 -romos -rostype load -romsrhs -writesol -visit

I suspect that it is something to do with the end or beginning of time window exactly coincides with the end of time steps.

@chldkdtn This happens when the offset option is on and the last time step is the only snapshot sample in the last sampling window, i.e. a sampling window is closed in the second-last time step. This leaves no snapshot vectors in the basis generator for position, velocity and energy. I think this issue is not introduced in this PR and it exists for a while but we don't have the coincidence to see this seg fault. I will fix this in this PR.

siuwuncheung commented 3 years ago

@siuwuncheung The regression tests all seem to be failing, and I suppose it is because of small differences (e.g. number of timesteps) due to round-off level changes caused by updating the FOM code from the master branch of Laghos. If so, that is fine, but can you please report whether you have verified this is the case? It is a lot of work to check every regression test output file to verify that the change is acceptable, so I think it is fine if just one person (usually the PR author) does this and reports the results.

@dylan-copeland Thanks for pointing that out. I will have a careful check on each regression test output.

chldkdtn commented 3 years ago

@siuwuncheung we need regression tests for ROM, RT instability problems. :)

chldkdtn commented 3 years ago

@siuwuncheung have you get a chance of looking into regression test failures?