Merge_beast #122

PaulOlyslager · 2024-03-12T14:31:25Z

added example for vppmchwt udsing composed operators
made it possible to redefine the geometry of a basis, this is necesarry to get the trace mesh into the basis.
added the posibility to use a general function as excitation
added dof_permutation for ndrefspace
corrected the ncross for rt and nd spaces.
added the possibility to construct general kernels (see composed operators, if other green functions are needed they can be added in a simple way (see implementation hh3dgreen))
-Traces are taken automatically with a geometric argument.
-Geometric support for traces added
-made sauterschwab momintegrals independent of normal by using permute_vertices
added a quadfunction argument to assemble and a kwargs argument to pass argements to lowest level of assembly. This makes it possible to add quadstrat functions for diffeertn type of operators. Programs using the original form will not fail.
-added linear combination of functionals
made it possible to set lhs to zero in equation
made the identity integrand more similar to the double integral integrands. (needed for composdoperator )
both test and trial cell are passed in local operator. This because in advance it is not known what the normals on both test and trial chart are.
made sure that sparsedict, densestorage and bandedstorage return exactly the same number
assembe_local_matched cannot be used for composed operators (see infinitely thin structures) (this needs to be wokred on further in the future)
added a zero element for the operator type, called ZeroOperator. multiplication with zero yields a zero operator.
added the possibility to write your operator in matrix form and to map it to a bilform
added the posibility to multiply a bilform with a diagonal matrix. to rescale the rows or colums in the matrix representaiton of this bilform.
replaced .vertices with verticeslist() because chart can be wrapped in tracechart or some other type of chart.
wrote function that assembly of equation can be done in two steps.
-test included for composed operator.
-corrected tests that use nxrt

test compare by taking limit (use two triangles)

a more general trace describtion is needed

implemented operators and trace, more general description needed

types of bases changed

test defaultquadstrad

included test HHHoperators in BEAST

non used functions commented

multi-trace approach

overwritten meshpoint to add normals

krcools · 2024-03-15T14:33:38Z

src/bases/bdm3dspace.jl

@@ -3,6 +3,7 @@ mutable struct BDM3DBasis{T,M,P} <: Space{T}
    fns::Vector{Vector{Shape{T}}}
    pos::Vector{P}
 end
+redefine_geometrie(basis::BDM3DBasis{T,M,P},geo::F) where {T,M,P,F}= BDM3DBasis{T,F,P}(geo,basis.fns,basis.pos)


Are you using this function?

Yes, to add the trace mesh to already defined bases, because the infinitisimal translation of the test function is done in the assembly

krcools · 2024-03-15T14:34:49Z

src/bases/local/laglocal.jl

@@ -150,7 +150,7 @@ function restrict(f::LagrangeRefSpace{T,1}, dom1, dom2) where T

    # for each point of the new domain
    for i in 1:D
-        v = dom2.vertices[i]
+        v = verticeslist(dom2)[i]


lets rename to vertices. Users might make assumptions on the return data structure if we add list.

verticeslist returns a list of vertices, the function vertices in compsciencemshes exists but it does an extra hcat to create a matrix, is this more expensive?

krcools · 2024-03-15T14:36:06Z

src/bases/local/ndlocal.jl

@@ -56,14 +55,36 @@ function restrict(ϕ::NDRefSpace{T}, dom1, dom2) where T

        u = carttobary(dom1, c)
        x = neighborhood(dom1, u)
-
+        σ = sign(dot(normalize(tangents(dom2,1)×tangents(dom2,2)),normal(dom2)))


We should probably add a function orientation for charts that returns +/-1 so that this potentially expensive and definitely ugly line can go.

totaly agree

krcools · 2024-03-15T14:38:02Z

src/bases/local/ndlocal.jl

        end
    end

    return Q
 end
+
+const _vert_perms_nd = [


did you synchronise with upstram and make your fork use the more general way of dealing with vertex permutations?

krcools · 2024-03-15T14:40:19Z

src/composedoperator/tracesimplex.jl

+TraceSimplex(simplex,direction::SVector)
+    simplex plus infinitesimal direction d, for each point p on the simplex, the trace is taken in the p-d direction
+"""
+struct TraceSimplex{T,U}


maybe ShiftedSimplex or TranslatedSimplex?

In essence they are shifted over an infinitisimal distance, maybe add this property in the name? for example InfTransSimplex ?

krcools · 2024-03-15T14:42:51Z

src/dyadicop.jl

@@ -35,7 +35,7 @@ end

 function assemble!(biop::DyadicOp, tfs::Space, bfs::Space, store,
    threading::Type{BEAST.Threading{:multi}};
-    quadstrat=defaultquadstrat(biop, tfs, bfs))
+    kwargs...)


Why is this better?

If we want further in the assemble! tree to add an argument, even for assembling one specific operator, we dont have to change all assemble functions. I think in a tree it is a good practise to always pas the kwargs argument, enabling us to add extra arguments for specific functions. in the leaves a kwargs argument is needed to catch unused argements intended for other leaves

krcools · 2024-03-15T14:44:43Z

src/solvers/itsolver.jl

@@ -91,8 +91,8 @@ function LinearAlgebra.mul!(y::AbstractVecOrMat, solver::GMRESSolver, x::Abstrac
    return y
 end

-LinearAlgebra.adjoint(A::GMRESSolver) = GMRESSolver(adjoint(A.linear_operator), A.maxiter, A.restart, A.abstol, A.reltol, A.verbose)
-LinearAlgebra.transpose(A::GMRESSolver) = GMRESSolver(transpose(A.linear_operator), A.maxiter, A.restart, A.abstol, A.reltol, A.verbose)
+LinearAlgebra.adjoint(A::GMRESSolver) = GMRESSolver(adjoint(A.linear_operator); A.left_preconditioner, A.maxiter, A.restart, A.abstol, A.reltol, A.verbose)


Why is this needed?

I just added some arguments because the function errored, I assume the function GMRESSolver was changed in the past

krcools · 2024-03-15T14:49:54Z

src/utils/variational.jl

+    out.coeff *= a
+    return out
+end
+function *(a::Union{Matrix,SMatrix},b::BilForm)


Too confusing I think. Users may expect pointwise multiplication.

krcools · 2024-03-15T14:51:22Z

src/utils/variational.jl

+end
+
+
+struct ZeroBilForm end


I think even the zero bilinear form needs to specify its domain Hilbert spaces

PaulOlyslager and others added 30 commits September 7, 2023 11:30

gitignore aangepast

0ed9669

MWncrossn implemented

9a6e0da

all operators implemented but not tested

790738e

ToDo: add all limits of the operators and write

e6aa985

test compare by taking limit (use two triangles)

last update 14 sept

a892426

update not working

d25271e

update

70dbfde

implemented operators, not tested

c566821

traces implemented, not tested

2fc4839

implemented operators and trace in a first manner

cab394b

a more general trace describtion is needed

implemented operators and trace, more general description needed

1335e2e

Merge pull request #1 from PaulOlyslager:trace-operators

9935473

implemented operators and trace, more general description needed

types of bases changed

3cfb1d4

Merge pull request #2 from PaulOlyslager:changing-types-of-bases

4b15721

types of bases changed

implemented operators in a general way

f8cb0a0

hhh debuged, TODO write compleet test,

a23cabb

test defaultquadstrad

debuged

b2cf8a0

created test HHHOperators

85c4172

included test HHHoperators in BEAST

03cfe78

Merge pull request #3 from PaulOlyslager:helmholtzhodge-operators

eac4d85

included test HHHoperators in BEAST

non used functions commented

2af25c7

Merge pull request #4 from PaulOlyslager:cleaning-up

2c17db7

non used functions commented

Merge pull request #5 from PaulOlyslager/implement-operators

4d58aff

design foundations of a general

d3e5502

multi-trace approach

work in progress

3f02d67

update

e383457

overwritten meshpoint to add normals

00cf973

Merge pull request #6 from PaulOlyslager:fix-local-operator

c97f948

overwritten meshpoint to add normals

trace operator in HHH created

d6147b9

normalorient implemented

b1378a3

PaulOlyslager and others added 19 commits March 11, 2024 15:01

cleanup composed operator

762deae

add documentation trace simplex and trace meshpoint

2e7cbda

matrix to bilform

2596a81

simplified trace and potential operator

d642dc2

update

460b352

Merge pull request #15 from PaulOlyslager/interactions

eb16173

update master to merge with beast

3ab7da4

remove files

907e204

update

a6aa7b7

update

2b42b55

update

18e36e0

removed exports, .gitignore adjusted

603607f

adjusted toml

68ffca2

updated vppmchwt to not rely on exports

682c69c

update to pass tests

65c1ed7

update vppmchwt

2a48cbc

farfield local for composed operator

9e4495b

update

f9bcea2

fix error restrict function

ddf51f4

krcools reviewed Mar 15, 2024

View reviewed changes

src/utils/variational.jl

end

struct ZeroBilForm end

Copy link

Owner

krcools Mar 15, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think even the zero bilinear form needs to specify its domain Hilbert spaces

PaulOlyslager added 2 commits November 12, 2024 13:37

update 1 before sync

41e7a0b

update 2 before sync

0e93c3f

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Merge_beast #122

Merge_beast #122

PaulOlyslager commented Mar 12, 2024

krcools Mar 15, 2024

PaulOlyslager May 22, 2024

krcools Mar 15, 2024

PaulOlyslager May 22, 2024

krcools Mar 15, 2024

PaulOlyslager May 22, 2024

krcools Mar 15, 2024

krcools Mar 15, 2024

PaulOlyslager May 22, 2024

krcools Mar 15, 2024

PaulOlyslager May 22, 2024

krcools Mar 15, 2024

PaulOlyslager May 22, 2024

krcools Mar 15, 2024

krcools Mar 15, 2024

Merge_beast #122

Are you sure you want to change the base?

Merge_beast #122

Conversation

PaulOlyslager commented Mar 12, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment