exaFOAM/foam-extend4.1-coherent-io
Archived
1
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Merge remote-tracking branch 'origin/hotfix/ggiSeparation+blockCoupledGgi' into nextRelease

Browse source 
Conflicts:
	src/OpenFOAM/meshes/polyMesh/polyPatches/constraint/ggi/ggiPolyPatch.H
This commit is contained in:
Dominik Christ 2013-07-08 12:05:49 +01:00
commit 8a70870771
14 changed files with 446 additions and 77 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
.gitignore vendored
View file

@ -101,4 +101,12 @@ src/lduSolvers/amg/amgPolicy/samgPolicy.H
src/lduSolvers/amg/amgPolicy/aamgPolicy.C
src/lduSolvers/amg/amgPolicy/aamgPolicy.H
# The following files are blacklisted because of a DMCA complaint by ANSYS.
src/lduSolvers/tools/PriorityArray.C
src/lduSolvers/tools/PriorityArray.H
src/lduSolvers/amg/amgPolicy/samgPolicy.C
src/lduSolvers/amg/amgPolicy/samgPolicy.H
src/lduSolvers/amg/amgPolicy/aamgPolicy.C
src/lduSolvers/amg/amgPolicy/aamgPolicy.H
# end-of-file

2
src/OpenFOAM/interpolations/GGIInterpolation/GGIInterpolation.H
View file

@ -95,7 +95,7 @@ public:
ClassName("GGIInterpolation");
//- Quick reject names
static const NamedEnum<quickReject, 3> quickRejectNames_;
static const NamedEnum<quickReject, 4> quickRejectNames_;
// Constructors

7
src/OpenFOAM/interpolations/GGIInterpolation/GGIInterpolationName.C
View file

@ -41,12 +41,15 @@ defineTypeNameAndDebug(Foam::GGIInterpolationName, 0);
template<>
const char*
Foam::NamedEnum<Foam::GGIInterpolationName::quickReject, 3>::names[] =
Foam::NamedEnum<Foam::GGIInterpolationName::quickReject, 4>::names[] =
{
"3DDistance",
"distance3D",
"AABB",
"bbOctree",
"nSquared"
};
const Foam::NamedEnum<Foam::GGIInterpolationName::quickReject, 4>
Foam::GGIInterpolationName::quickRejectNames_;
// ************************************************************************* //

69
src/OpenFOAM/matrices/blockLduMatrix/BlockLduMatrix/BlockLduMatrixATmul.C
View file

@ -64,15 +64,44 @@ void Foam::BlockLduMatrix<Type>::AmulCore
const unallocLabelList& u = lduAddr().upperAddr();
const unallocLabelList& l = lduAddr().lowerAddr();
const TypeCoeffField& Diag = this->diag();
const TypeCoeffField& Upper = this->upper();
// Diagonal multiplication, no indirection
multiply(Ax, Diag, x);
// Create multiplication function object
typename BlockCoeff<Type>::multiply mult;
// AmulCore must be additive to account for initialisation step
// in ldu interfaces. HJ, 6/Nov/2007
// Fixed by IC 19/Oct/2011
if (Diag.activeType() == blockCoeffBase::SCALAR)
{
const scalarTypeField& activeDiag = Diag.asScalar();
for (register label cellI = 0; cellI < u.size(); cellI++)
{
Ax[cellI] += mult(activeDiag[cellI], x[cellI]);
}
}
else if (Diag.activeType() == blockCoeffBase::LINEAR)
{
const linearTypeField& activeDiag = Diag.asLinear();
for (register label cellI = 0; cellI < u.size(); cellI++)
{
Ax[cellI] += mult(activeDiag[cellI], x[cellI]);
}
}
else if (Diag.activeType() == blockCoeffBase::SQUARE)
{
const squareTypeField& activeDiag = Diag.asSquare();
for (register label cellI = 0; cellI < u.size(); cellI++)
{
Ax[cellI] += mult(activeDiag[cellI], x[cellI]);
}
}
// Lower multiplication
if (symmetric())
@ -209,12 +238,40 @@ void Foam::BlockLduMatrix<Type>::TmulCore
const TypeCoeffField& Diag = this->diag();
const TypeCoeffField& Upper = this->upper();
// Diagonal multiplication, no indirection
multiply(Tx, Diag, x);
// Create multiplication function object
typename BlockCoeff<Type>::multiply mult;
// AmulCore must be additive to account for initialisation step
// in ldu interfaces. HJ, 6/Nov/2007
// Fixed by IC 19/Oct/2011
if (Diag.activeType() == blockCoeffBase::SCALAR)
{
const scalarTypeField& activeDiag = Diag.asScalar();
for (register label cellI = 0; cellI < u.size(); cellI++)
{
Tx[cellI] += mult(activeDiag[cellI], x[cellI]);
}
}
else if (Diag.activeType() == blockCoeffBase::LINEAR)
{
const linearTypeField& activeDiag = Diag.asLinear();
for (register label cellI = 0; cellI < u.size(); cellI++)
{
Tx[cellI] += mult(activeDiag[cellI], x[cellI]);
}
}
else if (Diag.activeType() == blockCoeffBase::SQUARE)
{
const squareTypeField& activeDiag = Diag.asSquare();
for (register label cellI = 0; cellI < u.size(); cellI++)
{
Tx[cellI] += mult(activeDiag[cellI].T(), x[cellI]);
}
}
// Upper multiplication
if (Upper.activeType() == blockCoeffBase::SCALAR)

163
src/OpenFOAM/matrices/blockLduMatrix/BlockLduMatrix/scalarBlockLduMatrix.C
View file

@ -248,23 +248,31 @@ void BlockLduMatrix<scalar>::AmulCore
const scalarField& x
) const
{
const scalarField& Diag = diag();
for (register label rowI = 0; rowI < x.size(); rowI++)
{
Ax[rowI] = Diag[rowI]*x[rowI];
}
// Note: pointer looping
const label* const __restrict__ U = lduAddr().upperAddr().begin();
const label* const __restrict__ L = lduAddr().lowerAddr().begin();
const scalar* const __restrict__ X = x.begin();
scalar* __restrict__ AX = Ax.begin();
if (thereIsDiag())
{
const scalar* const __restrict__ diagPtr = diag().begin();
register const label nCells = diag().size();
for (register label cell=0; cell<nCells; cell++)
{
// AmulCore must be additive to account for initialisation step
// in ldu interfaces. HJ, 6/Nov/2007
AX[cell] += diagPtr[cell]*X[cell];
}
}
if (symmetric())
{
if (thereIsUpper())
{
const scalar* const __restrict__ Upper = upper().begin();
const scalar* const __restrict__ X = x.begin();
scalar* __restrict__ AX = Ax.begin();
for (register label coeffI = 0; coeffI < upper().size(); coeffI++)
{
@ -272,13 +280,21 @@ void BlockLduMatrix<scalar>::AmulCore
AX[L[coeffI]] += Upper[coeffI]*X[U[coeffI]];
}
}
else
{
const scalar* const __restrict__ Lower = lower().begin();
for (register label coeffI = 0; coeffI < upper().size(); coeffI++)
{
AX[U[coeffI]] += Lower[coeffI]*X[L[coeffI]];
AX[L[coeffI]] += Lower[coeffI]*X[U[coeffI]];
}
}
}
else if (asymmetric())
{
const scalar* const __restrict__ Lower = lower().begin();
const scalar* const __restrict__ Upper = upper().begin();
const scalar* const __restrict__ X = x.begin();
scalar* __restrict__ AX = Ax.begin();
for (register label coeffI = 0; coeffI < upper().size(); coeffI++)
{
@ -296,23 +312,31 @@ void BlockLduMatrix<scalar>::TmulCore
const scalarField& x
) const
{
const scalarField& Diag = diag();
for (register label rowI = 0; rowI < x.size(); rowI++)
{
Tx[rowI] = Diag[rowI]*x[rowI];
}
// Note: pointer looping
const label* const __restrict__ U = lduAddr().upperAddr().begin();
const label* const __restrict__ L = lduAddr().lowerAddr().begin();
const scalar* const __restrict__ X = x.begin();
scalar* __restrict__ TX = Tx.begin();
if (thereIsDiag())
{
const scalar* const __restrict__ diagPtr = diag().begin();
register const label nCells = diag().size();
for (register label cell=0; cell<nCells; cell++)
{
// AmulCore must be additive to account for initialisation step
// in ldu interfaces. HJ, 6/Nov/2007
TX[cell] += diagPtr[cell]*X[cell];
}
}
if (symmetric())
{
if (thereIsUpper())
{
const scalar* const __restrict__ Upper = upper().begin();
const scalar* const __restrict__ X = x.begin();
scalar* __restrict__ TX = Tx.begin();
for (register label coeffI = 0; coeffI < upper().size(); coeffI++)
{
@ -320,13 +344,21 @@ void BlockLduMatrix<scalar>::TmulCore
TX[L[coeffI]] += Upper[coeffI]*X[U[coeffI]];
}
}
else
{
const scalar* const __restrict__ Lower = lower().begin();
for (register label coeffI = 0; coeffI < upper().size(); coeffI++)
{
TX[U[coeffI]] += Lower[coeffI]*X[L[coeffI]];
TX[L[coeffI]] += Lower[coeffI]*X[U[coeffI]];
}
}
}
else if (asymmetric())
{
const scalar* const __restrict__ Lower = lower().begin();
const scalar* const __restrict__ Upper = upper().begin();
const scalar* const __restrict__ X = x.begin();
scalar* __restrict__ TX = Tx.begin();
for (register label coeffI = 0; coeffI < upper().size(); coeffI++)
{
@ -368,9 +400,47 @@ tmp<scalarField> BlockLduMatrix<scalar>::H(const scalarField& x) const
const label* const __restrict__ U = lduAddr().upperAddr().begin();
const label* const __restrict__ L = lduAddr().lowerAddr().begin();
const scalar* const __restrict__ X = x.begin();
if (symmetric())
{
if (thereIsUpper())
{
const scalar* const __restrict__ Upper = upper().begin();
scalar* __restrict__ R = result.begin();
for (register label coeffI = 0; coeffI < upper().size(); coeffI++)
{
R[U[coeffI]] -= Upper[coeffI]*X[U[coeffI]];
}
for (register label coeffI = 0; coeffI < upper().size(); coeffI++)
{
R[L[coeffI]] -= Upper[coeffI]*X[L[coeffI]];
}
}
else if (thereIsLower())
{
const scalar* const __restrict__ Lower = lower().begin();
scalar* __restrict__ R = result.begin();
for (register label coeffI = 0; coeffI < upper().size(); coeffI++)
{
R[U[coeffI]] -= Lower[coeffI]*X[U[coeffI]];
}
for (register label coeffI = 0; coeffI < upper().size(); coeffI++)
{
R[L[coeffI]] -= Lower[coeffI]*X[L[coeffI]];
}
}
}
else
{
const scalar* const __restrict__ Lower = lower().begin();
const scalar* const __restrict__ Upper = upper().begin();
const scalar* const __restrict__ X = x.begin();
scalar* __restrict__ R = result.begin();
@ -384,6 +454,7 @@ tmp<scalarField> BlockLduMatrix<scalar>::H(const scalarField& x) const
R[L[coeffI]] -= Lower[coeffI]*X[L[coeffI]];
}
}
}
return tresult;
}
@ -396,6 +467,43 @@ tmp<scalarField> BlockLduMatrix<scalar>::faceH(const scalarField& x) const
scalarField& result = tresult();
if (thereIsUpper() || thereIsLower())
{
if (symmetric())
{
if (thereIsUpper())
{
// Note: pointer looping
const label* const __restrict__ U = lduAddr().upperAddr().begin();
const label* const __restrict__ L = lduAddr().lowerAddr().begin();
const scalar* const __restrict__ Upper = upper().begin();
const scalar* const __restrict__ X = x.begin();
scalar* __restrict__ R = result.begin();
for (register label coeffI = 0; coeffI < upper().size(); coeffI++)
{
R[coeffI] = Upper[coeffI]*(X[U[coeffI]] - X[L[coeffI]]);
}
}
else if (thereIsLower())
{
// Note: pointer looping
const label* const __restrict__ U = lduAddr().upperAddr().begin();
const label* const __restrict__ L = lduAddr().lowerAddr().begin();
const scalar* const __restrict__ Lower = lower().begin();
const scalar* const __restrict__ X = x.begin();
scalar* __restrict__ R = result.begin();
for (register label coeffI = 0; coeffI < upper().size(); coeffI++)
{
R[coeffI] = Lower[coeffI]*(X[U[coeffI]] - X[L[coeffI]]);
}
}
}
else
{
// Note: pointer looping
const label* const __restrict__ U = lduAddr().upperAddr().begin();
@ -412,6 +520,7 @@ tmp<scalarField> BlockLduMatrix<scalar>::faceH(const scalarField& x) const
R[coeffI] = Upper[coeffI]*X[U[coeffI]] - Lower[coeffI]*X[L[coeffI]];
}
}
}
return tresult;
}

1
src/OpenFOAM/matrices/lduMatrix/lduAddressing/lduInterfaceFields/ggiLduInterfaceField/ggiLduInterfaceField.C
View file

@ -69,4 +69,5 @@ void Foam::ggiLduInterfaceField::transformCoupleField
}
// ************************************************************************* //

2
src/OpenFOAM/meshes/polyMesh/polyPatches/constraint/cyclicGgi/cyclicGgiPolyPatch.C
View file

@ -257,7 +257,7 @@ const Foam::cyclicGgiPolyPatch& Foam::cyclicGgiPolyPatch::cyclicShadow() const
}
void Foam::cyclicGgiPolyPatch::calcTransforms()
void Foam::cyclicGgiPolyPatch::calcTransforms() const
{
if (active() && debug)
{

2
src/OpenFOAM/meshes/polyMesh/polyPatches/constraint/cyclicGgi/cyclicGgiPolyPatch.H
View file

@ -75,7 +75,7 @@ class cyclicGgiPolyPatch
//- Calculate cyclic transforms (rotation and translation)
// Virtual over-ride for base GGI patch. HJ, 14/Jan/2009
virtual void calcTransforms();
virtual void calcTransforms() const;
//- Check definition: angles and offsets
void checkDefinition() const;

28
src/OpenFOAM/meshes/polyMesh/polyPatches/constraint/ggi/ggiPolyPatch.C
View file

@ -202,8 +202,7 @@ void Foam::ggiPolyPatch::calcPatchToPatch() const
0, // Non-overlapping face tolerances
0, // HJ, 24/Oct/2008
true, // Rescale weighting factors. Bug fix, MB.
// ggiInterpolation::AABB
ggiInterpolation::BB_OCTREE // Octree search, MB.
reject_ // Quick rejection algorithm, default BB_OCTREE
);
// Abort immediately if uncovered faces are present and the option
@ -469,6 +468,7 @@ Foam::ggiPolyPatch::ggiPolyPatch
shadowName_("initializeMe"),
zoneName_("initializeMe"),
bridgeOverlap_(false),
reject_(ggiZoneInterpolation::BB_OCTREE),
shadowIndex_(-1),
zoneIndex_(-1),
patchToPatchPtr_(NULL),
@ -490,13 +490,15 @@ Foam::ggiPolyPatch::ggiPolyPatch
const polyBoundaryMesh& bm,
const word& shadowName,
const word& zoneName,
const bool bridgeOverlap
const bool bridgeOverlap,
const ggiZoneInterpolation::quickReject reject
)
:
coupledPolyPatch(name, size, start, index, bm),
shadowName_(shadowName),
zoneName_(zoneName),
bridgeOverlap_(bridgeOverlap),
reject_(reject),
shadowIndex_(-1),
zoneIndex_(-1),
patchToPatchPtr_(NULL),
@ -521,6 +523,7 @@ Foam::ggiPolyPatch::ggiPolyPatch
shadowName_(dict.lookup("shadowPatch")),
zoneName_(dict.lookup("zone")),
bridgeOverlap_(dict.lookup("bridgeOverlap")),
reject_(ggiZoneInterpolation::BB_OCTREE),
shadowIndex_(-1),
zoneIndex_(-1),
patchToPatchPtr_(NULL),
@ -530,7 +533,15 @@ Foam::ggiPolyPatch::ggiPolyPatch
localParallelPtr_(NULL),
receiveAddrPtr_(NULL),
sendAddrPtr_(NULL)
{}
{
if (dict.found("quickReject"))
{
reject_ = ggiZoneInterpolation::quickRejectNames_.read
(
dict.lookup("quickReject")
);
}
}
Foam::ggiPolyPatch::ggiPolyPatch
@ -543,6 +554,7 @@ Foam::ggiPolyPatch::ggiPolyPatch
shadowName_(pp.shadowName_),
zoneName_(pp.zoneName_),
bridgeOverlap_(pp.bridgeOverlap_),
reject_(pp.reject_),
shadowIndex_(-1),
zoneIndex_(-1),
patchToPatchPtr_(NULL),
@ -569,6 +581,7 @@ Foam::ggiPolyPatch::ggiPolyPatch
shadowName_(pp.shadowName_),
zoneName_(pp.zoneName_),
bridgeOverlap_(pp.bridgeOverlap_),
reject_(pp.reject_),
shadowIndex_(-1),
zoneIndex_(-1),
patchToPatchPtr_(NULL),
@ -742,6 +755,11 @@ void Foam::ggiPolyPatch::initAddressing()
// Calculate transforms for correct GGI cut
calcTransforms();
if (master())
{
shadow().calcTransforms();
}
// Force zone addressing and remote zone addressing
// (uses GGI interpolator)
zoneAddressing();
@ -847,7 +865,7 @@ void Foam::ggiPolyPatch::updateMesh()
}
void Foam::ggiPolyPatch::calcTransforms()
void Foam::ggiPolyPatch::calcTransforms() const
{
// Simplest mixing plane: no transform or separation. HJ, 24/Oct/2008
forwardT_.setSize(0);

13
src/OpenFOAM/meshes/polyMesh/polyPatches/constraint/ggi/ggiPolyPatch.H
View file

@ -73,6 +73,10 @@ class ggiPolyPatch
//- Use bridging to fix overlap error in interpolation
Switch bridgeOverlap_;
//- Quick reject algorithm
ggiZoneInterpolation::quickReject reject_;
// Demand-driven data
//- Shadow patch index. Delayed evaluation for construction
@ -214,7 +218,9 @@ public:
const polyBoundaryMesh& bm,
const word& shadowName,
const word& zoneName,
const bool bridgeOverlap
const bool bridgeOverlap,
const ggiZoneInterpolation::quickReject
reject = ggiZoneInterpolation::BB_OCTREE
);
//- Construct from dictionary
@ -307,6 +313,11 @@ public:
//- Return interpolation face zone
const faceZone& zone() const;
//- Is this the slave side?
bool slave() const
{
return !master();
}
// Interpolation data

29
src/finiteVolume/fields/fvPatchFields/constraint/cyclicGgi/cyclicGgiFvPatchField.C
View file

@ -289,6 +289,35 @@ void cyclicGgiFvPatchField<Type>::updateInterfaceMatrix
{}
template<class Type>
void cyclicGgiFvPatchField<Type>::initInterfaceMatrixUpdate
(
const Field<Type>& psiInternal,
Field<Type>& result,
const BlockLduMatrix<Type>& m,
const CoeffField<Type>& coeffs,
const Pstream::commsTypes commsType
) const
{
notImplemented("void cyclicGgiFvPatchField::initInterfaceMatrixUpdate"
" for block-coupled matrices of tensor types"
)
}
template<class Type>
void cyclicGgiFvPatchField<Type>::updateInterfaceMatrix
(
const Field<Type>&,
Field<Type>&,
const BlockLduMatrix<Type>&,
const CoeffField<Type>&,
const Pstream::commsTypes commsType
) const
{
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
} // End namespace Foam

19
src/finiteVolume/fields/fvPatchFields/constraint/cyclicGgi/cyclicGgiFvPatchField.H
View file

@ -168,6 +168,25 @@ public:
const bool switchToLhs
) const;
//- Initialise neighbour matrix update
virtual void initInterfaceMatrixUpdate
(
const Field<Type>& psiInternal,
Field<Type>& result,
const BlockLduMatrix<Type>& m,
const CoeffField<Type>& coeffs,
const Pstream::commsTypes commsType
) const;
//- Update result field based on interface functionality
virtual void updateInterfaceMatrix
(
const Field<Type>&,
Field<Type>&,
const BlockLduMatrix<Type>&,
const CoeffField<Type>&,
const Pstream::commsTypes commsType
) const;
//- GGI coupled interface functions

90
src/finiteVolume/fields/fvPatchFields/constraint/cyclicGgi/cyclicGgiFvPatchFields.C
View file

@ -37,12 +37,102 @@ Contributor:
#include "cyclicGgiFvPatchFields.H"
#include "addToRunTimeSelectionTable.H"
#include "volFields.H"
#include "scalarCoeffField.H"
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
namespace Foam
{
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
template<>
void cyclicGgiFvPatchField<scalar>::initInterfaceMatrixUpdate
(
const Field<scalar>& psiInternal,
Field<scalar>& result,
const BlockLduMatrix<scalar>& m,
const CoeffField<scalar>& coeffs,
const Pstream::commsTypes commsType
) const
{
// Communication is allowed either before or after processor
// patch comms. HJ, 11/Jul/2011
// Get shadow face-cells and assemble shadow field
const unallocLabelList& sfc = cyclicGgiPatch_.shadow().faceCells();
scalarField sField(sfc.size());
forAll (sField, i)
{
sField[i] = psiInternal[sfc[i]];
}
// Note: scalar interpolate does not get a transform, so this is safe
// HJ, 12/Jan/2009
scalarField pnf = cyclicGgiPatch_.interpolate(sField);
// Multiply the field by coefficients and add into the result
const unallocLabelList& fc = cyclicGgiPatch_.faceCells();
forAll(fc, elemI)
{
result[fc[elemI]] -= coeffs[elemI]*pnf[elemI];
};
}
template<>
void cyclicGgiFvPatchField<vector>::initInterfaceMatrixUpdate
(
const Field<vector>& psiInternal,
Field<vector>& result,
const BlockLduMatrix<vector>& m,
const CoeffField<vector>& coeffs,
const Pstream::commsTypes commsType
) const
{
// Communication is allowed either before or after processor
// patch comms. HJ, 11/Jul/2011
// Get shadow face-cells and assemble shadow field
const unallocLabelList& sfc = cyclicGgiPatch_.shadow().faceCells();
Field<vector> sField(sfc.size());
forAll (sField, i)
{
sField[i] = psiInternal[sfc[i]];
}
// Transformation is handled in interpolation. HJ, 7/Jan/2009
Field<vector> pnf = cyclicGgiPatch_.interpolate(sField);
if (coeffs.activeType() == blockCoeffBase::SCALAR)
{
pnf = coeffs.asScalar() * pnf;
}
else if (coeffs.activeType() == blockCoeffBase::LINEAR)
{
pnf = cmptMultiply(coeffs.asLinear(), pnf);
}
else if (coeffs.activeType() == blockCoeffBase::SQUARE)
{
pnf = coeffs.asSquare() & pnf;
}
// Multiply the field by coefficients and add into the result
const unallocLabelList& fc = cyclicGgiPatch_.faceCells();
// Multiply the field by coefficients and add into the result
forAll(fc, elemI)
{
result[fc[elemI]] -= pnf[elemI];
}
}
// * * * * * * * * * * * * * * Static Data Members * * * * * * * * * * * * * //
makePatchFields(cyclicGgi);

24
src/finiteVolume/fields/fvPatchFields/constraint/cyclicGgi/cyclicGgiFvPatchFields.H
View file

@ -51,6 +51,30 @@ SourceFiles
namespace Foam
{
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
template<>
void cyclicGgiFvPatchField<scalar>::initInterfaceMatrixUpdate
(
const Field<scalar>& psiInternal,
Field<scalar>& result,
const BlockLduMatrix<scalar>& m,
const CoeffField<scalar>& coeffs,
const Pstream::commsTypes commsType
) const;
template<>
void cyclicGgiFvPatchField<vector>::initInterfaceMatrixUpdate
(
const Field<vector>& psiInternal,
Field<vector>& result,
const BlockLduMatrix<vector>& m,
const CoeffField<vector>& coeffs,
const Pstream::commsTypes commsType
) const;
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
makePatchTypeFieldTypedefs(cyclicGgi)