Make PyWENO kernels work with Fortran SharpClaw code about pyclaw HOT 9 CLOSED

Happy to help with this one... is there a test case for this?

from pyclaw.

Hi Matt,

Sorry for the slow reply; I was on vacation when you chimed in. There are
several test cases for sharpclaw, so the first thing would be to get a
PyWENO-generated fifth-order kernel working and make sure the tests pass.
Then it should be easy to add other orders. Here's what I think needs to
be done:

1. Put the PyWENO-generated code into a new fortran module that does what
   reconstruct.py does (i.e., takes q and computes ql, qr -- see also
   reconstruct.f90 in pyclaw/src/fortran/1d/sharpclaw).
2. Modify flux1.f90 (in the same directory) to use the new module instead of
   the old one.

I think we'd want to change the code significantly from what PyWENO
generates in order to make it more efficient (or better yet, we should add
PyWENO routines that generate more efficient, though probably less readable
code as an option). But I think the first thing to do is just get it
working.

-David

On Tue, Aug 2, 2011 at 4:38 AM, memmett <
[email protected]>wrote:

Happy to help with this one... is there a test case for this?

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

from pyclaw.

Alrighty, I'll get right on it.

Regarding making it more efficient: PyWENO allows some flexibility in combining several steps (smoothness, weights, reconstruction) into one function call. By combining everything into one function, there might be some redundant computation, but much less data movement. After I get things working I'll try this out.

Matt

from pyclaw.

Hi David,

I'm going to start with the weno5 routine in flux1.f90. Can the other weno5_* routines be re-written to call the plain weno5 routine?

Matt

from pyclaw.

That sounds good. Let's not worry about the other reconstructions (wave,
characteristic, etc.) for now. I want to do them eventually, but we don't
need them for the paper.

-David

On Mon, Aug 15, 2011 at 8:51 PM, memmett <
[email protected]>wrote:

Hi David,

I'm going to start with the weno5 routine in flux1.f90. Can the other
weno5_* routines be re-written to call the plain weno5 routine?

Matt

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

from pyclaw.

Do you have a preference for how to select the WENO order? How about a 'weno_order' attribute of SharpClawSolver that defaults to 5?

from pyclaw.

That sounds good to me.

David Ketcheson

Sent from my iphone

On Aug 19, 2011, at 6:07 PM, [email protected] wrote:

Do you have a preference for how to select the WENO order? How about a 'weno_order' attribute of SharpClawSolver that defaults to 5?

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

from pyclaw.

Pull request submitted... will close this issue when the request gets merged.

from pyclaw.

Complete as of cd5daf3.

from pyclaw.