Create sorting routine for ids.
Update all inputs. Look at that in more details, ie with respect to control methods. When do we do update of inputs?  Handle input update for Nodes and Materials. 
Define default cluster in /etc/cielo.rc
VelocityExtrude: debug 
FindParam(vx) then DeleteParam("vx")? Should vx be a Vec? Can we find a way of partionting vx with partition, and staying in parallel?
Update control to get new inputs, as well as parallel diagnostic

Finish diagnostic_core.m
Create sorting routine for ids.
WriteData for OutputDiagnostic
OutputDiagnostic not working in 3d. (wrong gsize)
Finish thermal.m solution, and thermal.cpp solutoin. bukld thermal element.
Check that gaussian points get erased everywhere!
Check that GetJacobianDeterminant is computed correctly. 
Lighten up on parameters when they have been used.
Create new framework for outputing parallel results.


Look at: 
ProcessParamsx
OutputThermal
CreateFemModel
Dakota patch
Finish root, suffix, isnumeric in shared/String. Then, do partition recover in SpawnCore.

ismodelresultsconsistent: do it with results field now. suppress vx,vy,vz,vel,temperature,melting,transient_results, etc ...
lighten up the parameters, make a routine.
check that ismodelselfconsistent checks concurrency is 0 when running in parallel!
Fix problem that dakota does not MPI_Finalize correctly! 
only min_vel is processed!
Make sure inputs are updated homogeneously: ex: if updating thickness, surface and bed also need to be updated! this could get complicated, for 
ex, if an object depends only on surface, but thickness gets updated, then the object should still pick up the surface change (if on a bedrock for example).think 
about that quite hard.
Debug Dakota new code. Then think about systemtizing ModelProcessor! It's getting out of hands + lighten up on parameters! 
Write qmu doc, change names may be?

Answer Jesse, Bob.







Look at issm_options.m4, homogeneize! 
Look at petsc, why doesn't it compile?
Homogeneize the archive.







Look at issm_options.m4, homogeneize! 
Look at petsc, why doesn't it compile?
Homogeneize the archive.







Look at issm_options.m4, homogeneize! 
Look at petsc, why doesn't it compile?
Homogeneize the archive.







Look at issm_options.m4, homogeneize! 
Look at petsc, why doesn't it compile?
Homogeneize the archive.


Dakota: put misfit in: DakotaResponses is ok But now, we need parameters fit and u_g_obs put in in QmuCreateParameters. Then Misfitx in DakotaResponses should be ok.


Get Matlab 7.8: version 4.2.3
Use gprof to figure out where we are losing out time. 
test

test commit
test2
