Implicit time stepping about pyclaw HOT 8 OPEN

Matteo, is this related to #22? Is this something you need help with or you are putting up for personal tracking?

from pyclaw.

Hi Aron,
no #22 is related to Jed implementation of implicit TS with PETSc. Issue #33
is different: we want to implement our own implicit TS and use petsc to
solve the nonlinear system.
...but sure your are an expert of PetClaw so your help is precious!

On 20 June 2011 21:23, ahmadia <
[email protected]>wrote:

Matteo, is this related to #22? Is this something you need help with or
you are putting up for personal tracking?

Reply to this email directly or view it on GitHub:
#33 (comment)

Matteo Parsani

from pyclaw.

Starting date: tomorrow, July 2, 2011

from pyclaw.

Hi Matteo,

I don't know nearly enough about the functions you are calling/modifying to
give you any advice out of hand, but if you think it's worth trying to
explain the mathematics and computations to me, I will do my best to advise
your refactoring strategy :)

Aron

On Sat, Jul 2, 2011 at 3:13 PM, mparsani <
[email protected]>wrote:

The implementation of the subroutine that construct the nonlinear function
requires:

• implementation of a new function that "replaces/modifies" step1/2.f for
  the classic solver
  • implementation of a new function that "replaces/modifies" flux1.f90 for
    the sharpclaw solver

Those two functions will call the Riemann solver and manipulate the
fluctuations to get the nonlinear function. This will be passed back to
ClawSolver1D/2D or SharpClawSolver1D/2D (in python) which will call the
PETSc nonlinear solver.

Is this okay for all the developers? Do you have a better solution?

Reply to this email directly or view it on GitHub:
#33 (comment)

from pyclaw.

SharpClawSolver: uses the method of lines. Thus, we already have all the functions to construct the nonlinear function.
Classic implicit : the step function in clawpack.py must be modified when the implicit LW is used. In addition, Classic1.f can not be used.

from pyclaw.

I think it makes sense to extend SharpClaw in this case then.

A

On Sat, Jul 2, 2011 at 4:55 PM, mparsani <
[email protected]>wrote:

SharpClawSolver: uses the method of lines. Thus, we already have all the
functions to construct the nonlinear function.
Classic implicit : the step function in clawpack.py must be modified when
the implicit LW is used. In addition, Classic1.f can not be used.

Reply to this email directly or view it on GitHub:
#33 (comment)

from pyclaw.

Yes. I'll start with SharpClaw.

Thanks.

Matteo

On 2 July 2011 17:22, ahmadia <
[email protected]>wrote:

I think it makes sense to extend SharpClaw in this case then.

A

On Sat, Jul 2, 2011 at 4:55 PM, mparsani <
[email protected]>wrote:

SharpClawSolver: uses the method of lines. Thus, we already have all the
functions to construct the nonlinear function.
Classic implicit : the step function in clawpack.py must be modified when
the implicit LW is used. In addition, Classic1.f can not be used.

Reply to this email directly or view it on GitHub:
#33 (comment)

Reply to this email directly or view it on GitHub:
#33 (comment)

Matteo Parsani

from pyclaw.

Last message for today, sorry: I did not take into account that sharpclaw has not been tested thoroughly with mapped grids. Thus, since for the paper we want an Euler flow over a round cylinder, I'll start with the implicit LW.

Best Regards,
Matteo

from pyclaw.