modify global.h to turn on adaptivity

I've taken these minutes and moved them over to trac: http://dis.cs.uiuc.edu/trac/cpsd/

2006-10-13

Minutes:

• parallel
  • Aaron is re-arranging the makefiles so that you can build shared/static libraries.
  • need to fix the stream viewer on the macs
  • Need to update our driver routine to better match their main routine
  • test non-adaptivity for one vp, get that working.
• switch output based streaming output to patch level streaming output. Possible to checkpoint from output pach files?

2006-10-06

Minutes:

• Parallel progress
  • fixed the ascii output issues. (took longer than expected)
  • eventually, we need a class or object that holds all the streams and file pointers to open files.
  • CPSDPAR macro - what does it do?
    • code under this macro should probably need to be removed

2006-09-29

Minutes:

• Parallel progress
  • parallel visualization programs work
  • still having problems with global output varaibles, gOutput_Dir may need to be changed to something more general in the future.
• Big discussion on how to merge the branches
  • Rob is going to checkpoint everything into a parallel checkpoint branch, so we can continue to be productive while the meshing goes on
  • new_geom merges with moving boundary
  • moving_boundary
• changes to the serial code, optimizations
  • big speedups due to optimization flags
  • redesigning loops, trimming inner loops
  • Need to work out element triangle local numbering with chao hang after the meeting
  • Question for parallel work: Is it ok to pre-compute things for element solving?
    • answer- yes, if they don't need to be migrated
• Rob is going to set up a trac repository, tell everyone else about it.
• Geometry by Dr. Erikson
  • causal: norm(grad(t_)) <= 1 (after scaling)
  • 1 progressive - tri alpha is causal for all t_1
  • 2 progressive - tri alpha, beta is causal for all

2006-09-22

• Discussion over how to structure the meeting.
  • possibly
• discussion of parallel
  • progress? Still having trouble visualizing/debugging
  • need to communicate better with the physics people
  • Other: need code that can stitch the slab files back together.
  • Aaron+Rob - have output files in output/ go into output.0/ etc.
  • Aaron+Rob - find out why code isn't accepting, go inside physics code.
  • Rob - work closer with Aaron
  • 205 H talbot (in corner)

2006-09=15

• parallel
  • code runs, no obvious errors, need to check errors
  • when run on multiple processors, quits like everything is ok
  • had to make assumptions, what happens if you update a triangle on another processor that doesn't exist?
    • do nothing for now
    • might need to worry about edge flips
    • Aaron will email Zhao Huang.
  • No refinement when run on smaller numbers of VPs - probably indicates an error
  • Need a way to visualize the solution. We need an approved computer that is running windows
    • need to test both my laptops.
    • for now, send the file to someone else, ask them to visualize it.
    • need to work more closely with Resin
  • Need to look up how triangles get deleted.
• restart
  • can't go above characteristic lines.
  • if max((t2-t1)/(x2-x1)) < D, then safe to restart.
• sequential performance tuning
  • plan ot switch to blitz++ within the week
  • 400% facter than TNT< by guestimates.
  • run the code through gprof.
• meshing issues

2006-09-01

Minutes

• Aaron thinks he's finished with his changes that make the ghost edge flip
• problems with passing functions pointers, Aaron and ??? need to hash out the details.
• decided that edge split should block, just to start with.
  • need to look into more than one refinement wave running at once.
• Rob will be working with adaptivity.
• progress on the moving boundary part of the meshing.
  • degree of vertex: number of edges connected to a vertex ?

2006-08-25

Minutes:

• Aaron is working on a bug where the terrain refine wants to refine a ghost node, needs to be propagated to the proper chunk

within the CPSD, there are the following CS concepts.

• visualization
• space time meshing
• parallelization