Comment (by helvi witek):

Adding my original report to Zach and our respective replies:

Report on Baikal and BaikalVacuum (13 May 2020)

• Two thorns "Baikal" (Einstein's equations coupled to Tmn) and "BaikalVacuum" (Einstein's equations in vacuum)
• very clean and readable implementation, both the the jupyter notebook and the auto-generated code.
• exceptionally well documented
• benchmarked against McLachlan (and Lean test is underway): reproduces standard qc0 run (BBH with 1 orbit before merger)

Baikal (13 May 2020):

• README:

  • "...in parfile_examples/" should be "... in par/" to be consistent with source code structure

• interface.ccl

  • definition of BSSN grid functions: you might consider using tags for the tensortype and tensorweight.
    (Q: how does it impact the evolution or is it only information?).
    Doing so would require to split the evolution variables into groups of same type and weight.

• param.ccl:

  • [CHECK:] no floor_value for conformal factor or lapse -- how is the zero at the puncture treated?

• schedule.ccl:

  • Why do you store the ADMBase gridfunctions (again) in Baikal? This is done in ADMBase/schedule.ccl, so why is it necessary here?
  • [CHECK:] BSSN-to-ADM conversion before calculating Ricci and rhs. Why?

• src/Symmetry_registration_oldCartGrid3D.c:

  • [CORRECTION:] default symmetry setting is hardwired that of a scalar in every tensor component.
    This should be corrected, otherwise using symmetries to reduce run costs might yield incorrect results.
    For example, gammax has symmetry (-1,1,1).

• src/BSSN_to_ADM.c

  • [CHECK:] trafo is $\gamma_{xx} = \frac{1}{cf^2} (\tilde{\gamma}_{xx}+1 )$ (why "+1"?) and same for yy and zz

    --
    Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2299/proposal-add-baikaletk-to-the-toolkit