- 08 Aug, 2016 40 commits
#820 Start introducing a dev method to properly write on disk the global linear algebra to be compared with Freefem (it's solely for dev purposes).
#820 I have required higher precision requirements in Gmres to get more iterations; it works well (so dH is done!).
#820 Refine indexes used to tag iteration by splitting explicitly differential/non differential case (useful for debug comparison with Freefem).
#820 Add missing ZeroEntries(). It was also the cause of the 2. factor that bothered me so much yesterday...
#820 SolidDeltaResidual: computation is now closer to what we really want (there is only a factor missing). However a factor 2. introduced as a hack is still there; this will require investigation (it won't work in all likelihood in later iterations).
#820 Empirically correct the factor 2. I currently get on my dev operator; this will require further investigation later!
#820 Assemble of SolidDeltaResidual runs correctly (but does not yet compute what is intended: it's still the content from SecondPiolaKirchhoffStressTensor...). There is currently a factor 2. discrepancy with same quantity computed for dev purpose in Freefem.
#820 Assemble of SolidDeltaResidual runs correctly (but does not yet compute what is intended: it's still the content from SecondPiolaKirchhoffStressTensor...).
#820 Add midpoint position in implicit step fluid (at the moment we assume midpoint is used throughout the model; a static_assert checks it).
#820 Add VariableHolder in SolidVariationalFormulation (and stop redefining there a new Solid object).
#820 SolidDeltaResidual: incorporate some methods that were in hyperelasticity policy in SecondPiolKirchhoffStressTensor.
#820 Introduce an operator SolidDeltaResidual that should compute the 'deltaResidualFSIMiddlePoint' of dH.edp. At the moment it's just a trimmed down version of the SecondPiolaKirchhoffStressTensor.
#820 Fix a bug introduced yesterday: the matrix used in dH was not the right one (Dirichlet contribution was still there; it didn't appear on numerical results of first iteration but could have been tricky to find in later ones).