Add option to interpolate data on faces linearly in the tangential di…

[asalmgren]

…rection -- previously we only did piecewise constant within the face

Summary

Additional background

Checklist

The proposed changes:

• fix a bug or incorrect behavior in AMReX
• [X ] add new capabilities to AMReX
• changes answers in the test suite to more than roundoff level
• are likely to significantly affect the results of downstream AMReX users
• include documentation in the code and/or rst files, if appropriate

[WeiqunZhang]

Suggested change

	// We don't need to worry about face-based domain because this is only used in the tangential interpolation
	Box per_grown_domain = m_crse_geom.Domain();
	for (int dim = 0; dim < AMREX_SPACEDIM; dim++) {
	if (m_crse_geom.isPeriodic(dim)) {
	per_grown_domain.grow(dim,1);
	}
	}

Don't see it being used anywhere.

[WeiqunZhang]

Suggested change

ng[i] = 0;

ng is no longer being used at this point.

[WeiqunZhang]

Is the testing necessary? The CoarsenBox function grows the coarsened box by one cell. If the testing is necessary, it might be safer to test with > and < rather than !=.

[asalmgren]

I should have made this a member of the AMReX_InterpFaceRegister class -- it is used in the calls to interp_face_reg. Not sure whether it also needs to be a member of tags?

…

On Tue, Mar 12, 2024 at 8:43 AM Weiqun Zhang ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In Src/AmrCore/AMReX_InterpFaceRegister.cpp <#3794 (comment)>: > + // We don't need to worry about face-based domain because this is only used in the tangential interpolation + Box per_grown_domain = m_crse_geom.Domain(); + for (int dim = 0; dim < AMREX_SPACEDIM; dim++) { + if (m_crse_geom.isPeriodic(dim)) { + per_grown_domain.grow(dim,1); + } + } + ⬇️ Suggested change - // We don't need to worry about face-based domain because this is only used in the tangential interpolation - Box per_grown_domain = m_crse_geom.Domain(); - for (int dim = 0; dim < AMREX_SPACEDIM; dim++) { - if (m_crse_geom.isPeriodic(dim)) { - per_grown_domain.grow(dim,1); - } - } - Don't see it being used anywhere. — Reply to this email directly, view it on GitHub <#3794 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACRE6YR2GDP4MVBP4NLMBZLYX4PC5AVCNFSM6AAAAABEPO2752VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMYTSMZRGQ3DCNBXGA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

-- Ann Almgren Senior Scientist; Dept. Head, Applied Mathematics Pronouns: she/her/hers

[asalmgren]

The idea is for this to work without having filled ghost cell values of the coarse data outside the domain. More generally we would want to enforce the coarse data was filled there

…

On Tue, Mar 12, 2024 at 9:19 AM Weiqun Zhang ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In Src/AmrCore/AMReX_Interp_C.H <#3794 (comment)>: > +#else + int cj = amrex::coarsen(j, ratio[1]); +#endif + +#if (AMREX_SPACEDIM == 3) + int ck = amrex::coarsen(k, ratio[2]); +#else + int ck = 0; +#endif + + if (dim == 0 && ci*ratio[0] == i) { + // Check solve mask to ensure we don't overwrite valid fine data. + if (!mask || mask(ci, cj, ck, n)) { + fine(i, j, k, n) = crse(ci, cj, ck, n); +#if (AMREX_SPACEDIM >= 2) + if (cj != per_grown_domain.smallEnd(1) && cj != per_grown_domain.bigEnd(1) && ratio[1] > 1) { Is the testing necessary? The CoarsenBox function grows the coarsened box by one cell. If the testing is necessary, it might be safer to test with > and < rather than !=. — Reply to this email directly, view it on GitHub <#3794 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACRE6YTYFRKK6L4KQYRSRGTYX4TH7AVCNFSM6AAAAABEPO2752VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMYTSMZRGU2TCMJQG4> . You are receiving this because you authored the thread.Message ID: ***@***.***>

-- Ann Almgren Senior Scientist; Dept. Head, Applied Mathematics Pronouns: she/her/hers

[WeiqunZhang]

The idea is for this to work without having filled ghost cell values of the coarse data outside the domain. More generally we would want to enforce the coarse data was filled there

Not sure I understand. The current code with the checking has already relied on that the coarser data outside periodic boundary being filled. Otherwise the check should be using normal domain box, not per_grown_domain, unless the idea is we only fill one ghost cell of the coarse data, not more than one.