Feature single precision solve type

This commit is contained in:
Mattijs Janssens 2019-11-19 11:10:07 +00:00 committed by Andrew Heather
parent f2eb3e1cee
commit 2d080ff331
12 changed files with 293 additions and 38 deletions

View File

@ -24,6 +24,10 @@ License
Description
Simple field tests
Test use of Kahan/Neumaier to extend precision for when running SPDP
mode. Conclusion is that it is easier/quicker to run these summation
loops as double precision (i.e. solveScalar).
\*---------------------------------------------------------------------------*/
#include "primitiveFields.H"
@ -32,6 +36,116 @@ Description
using namespace Foam;
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
template<class Type, class CombineOp, class ResultType>
void sumNeumaier
(
const UList<Type>& vals,
const CombineOp& cop,
ResultType& result
)
{
// Neumaier version of Kahan
ResultType sum = Zero;
ResultType c = Zero;
for (const Type& vali : vals)
{
ResultType val;
cop(val, vali);
const ResultType t = sum + val;
for
(
direction cmpt = 0;
cmpt < pTraits<ResultType>::nComponents;
cmpt++
)
{
if (mag(sum[cmpt]) >= mag(val[cmpt]))
{
// If sum is bigger, low-order digits of input[i] are lost.
c[cmpt] += (sum[cmpt] - t[cmpt]) + val[cmpt];
}
else
{
// Else low-order digits of sum are lost.
c[cmpt] += (val[cmpt] - t[cmpt]) + sum[cmpt];
}
}
sum = t;
}
result = sum + c;
}
template<class CombineOp, class ResultType>
void sumNeumaier
(
const UList<scalar>& vals,
const CombineOp& cop,
ResultType& result
)
{
// Neumaier version of Kahan
ResultType sum = Zero;
ResultType c = Zero;
for (const scalar vali : vals)
{
ResultType val;
cop(val, vali);
const ResultType t = sum + val;
if (mag(sum) >= mag(val))
{
// If sum is bigger, low-order digits of input[i] are lost.
c += (sum - t) + val;
}
else
{
// Else low-order digits of sum are lost.
c += (val - t) + sum;
}
sum = t;
}
result = sum + c;
}
template<class Type>
Type mySum(const UList<Type>& f)
{
typedef typename Foam::typeOfSolve<Type>::type solveType;
solveType Sum = Zero;
if (f.size())
{
sumNeumaier(f, eqOp<solveType>(), Sum);
}
return Type(Sum);
}
//- The sumSqr always adds only positive numbers. Here there can never be any
// cancellation of truncation errors.
template<class Type>
typename outerProduct1<Type>::type mySumSqr(const UList<Type>& f)
{
typedef typename outerProduct1<solveScalar>::type prodType;
prodType result = Zero;
if (f.size())
{
sumNeumaier(f, eqSqrOp<prodType>(), result);
}
return result;
}
// Main program:
int main(int argc, char *argv[])
@ -63,6 +177,71 @@ int main(int argc, char *argv[])
Info<< "negated: " << lfield << nl;
// Summation (compile in SPDP)
// ~~~~~~~~~~~~~~~~~~~~~~~~~~~
Pout.precision(16);
Sout.precision(16);
const scalar SMALLS(1e-6);
const vector SMALLV(SMALLS, SMALLS, SMALLS);
const scalar GREATS(1e6);
const vector GREATV(GREATS, GREATS, GREATS);
// scalarField summation
{
scalarField sfield(10, SMALLS);
sfield[8] = GREATS;
sfield[9] = -sfield[8];
Info<< "scalarField:" << sfield.size() << nl
<< " sum :" << sum(sfield) << nl
<< " corrected:" << mySum(sfield) << endl;
}
// vectorField summation
{
vectorField vfield(10, SMALLV);
vfield[8] = GREATV;
vfield[9] = -vfield[8];
Info<< "vectorField:" << vfield.size() << nl
<< " sum :" << sum(vfield) << nl
<< " corrected:" << mySum(vfield) << endl;
}
// sphericalTensorField summation
{
sphericalTensorField tfield(10, SMALLS);
tfield[8] = GREATS;
tfield[9] = -tfield[8];
Info<< "sphericalTensorField:" << tfield.size() << nl
<< " sum :" << sum(tfield) << nl
<< " corrected:" << mySum(tfield) << endl;
}
// symmTensorField summation
{
symmTensorField tfield(10, SMALLS*symmTensor::I);
tfield[8] = GREATS*symmTensor::I;
tfield[9] = -tfield[8];
Info<< "symmTensorField:" << tfield.size() << nl
<< " sum :" << sum(tfield) << nl
<< " corrected:" << mySum(tfield) << endl;
}
// tensorField summation
{
tensorField tfield(10, SMALLS*tensor::I);
tfield[8] = GREATS*tensor::I;
tfield[9] = -tfield[8];
Info<< "tensorField:" << tfield.size() << nl
<< " sum :" << sum(tfield) << nl
<< " corrected:" << mySum(tfield) << endl;
}
return 0;
}

View File

@ -390,14 +390,16 @@ TMP_UNARY_FUNCTION(Type, min)
template<class Type>
Type sum(const UList<Type>& f)
{
Type Sum = Zero;
typedef typename Foam::typeOfSolve<Type>::type solveType;
solveType Sum = Zero;
if (f.size())
{
TFOR_ALL_S_OP_F(Type, Sum, +=, Type, f)
TFOR_ALL_S_OP_FUNC_F(solveType, Sum, +=, solveType, Type, f)
}
return Sum;
return Type(Sum);
}
TMP_UNARY_FUNCTION(Type, sum)

View File

@ -30,6 +30,7 @@ Description
\*---------------------------------------------------------------------------*/
#include "primitiveMesh.H"
#include "PrecisionAdaptor.H"
// * * * * * * * * * * * * * Private Member Functions * * * * * * * * * * * //
@ -75,10 +76,17 @@ void Foam::primitiveMesh::makeCellCentresAndVols
(
const vectorField& fCtrs,
const vectorField& fAreas,
vectorField& cellCtrs,
scalarField& cellVols
vectorField& cellCtrs_s,
scalarField& cellVols_s
) const
{
typedef Vector<solveScalar> solveVector;
PrecisionAdaptor<solveVector, vector> tcellCtrs(cellCtrs_s);
Field<solveVector>& cellCtrs = tcellCtrs.ref();
PrecisionAdaptor<solveScalar, scalar> tcellVols(cellVols_s);
Field<solveScalar>& cellVols = tcellVols.ref();
// Clear the fields for accumulation
cellCtrs = Zero;
cellVols = 0.0;
@ -89,18 +97,18 @@ void Foam::primitiveMesh::makeCellCentresAndVols
// first estimate the approximate cell centre as the average of
// face centres
vectorField cEst(nCells(), Zero);
Field<solveVector> cEst(nCells(), Zero);
labelField nCellFaces(nCells(), Zero);
forAll(own, facei)
{
cEst[own[facei]] += fCtrs[facei];
cEst[own[facei]] += solveVector(fCtrs[facei]);
++nCellFaces[own[facei]];
}
forAll(nei, facei)
{
cEst[nei[facei]] += fCtrs[facei];
cEst[nei[facei]] += solveVector(fCtrs[facei]);
++nCellFaces[nei[facei]];
}
@ -111,12 +119,15 @@ void Foam::primitiveMesh::makeCellCentresAndVols
forAll(own, facei)
{
const solveVector fc(fCtrs[facei]);
const solveVector fA(fAreas[facei]);
// Calculate 3*face-pyramid volume
scalar pyr3Vol =
fAreas[facei] & (fCtrs[facei] - cEst[own[facei]]);
solveScalar pyr3Vol =
fA & (fc - cEst[own[facei]]);
// Calculate face-pyramid centre
vector pc = (3.0/4.0)*fCtrs[facei] + (1.0/4.0)*cEst[own[facei]];
solveVector pc = (3.0/4.0)*fc + (1.0/4.0)*cEst[own[facei]];
// Accumulate volume-weighted face-pyramid centre
cellCtrs[own[facei]] += pyr3Vol*pc;
@ -127,12 +138,15 @@ void Foam::primitiveMesh::makeCellCentresAndVols
forAll(nei, facei)
{
const solveVector fc(fCtrs[facei]);
const solveVector fA(fAreas[facei]);
// Calculate 3*face-pyramid volume
scalar pyr3Vol =
fAreas[facei] & (cEst[nei[facei]] - fCtrs[facei]);
solveScalar pyr3Vol =
fA & (cEst[nei[facei]] - fc);
// Calculate face-pyramid centre
vector pc = (3.0/4.0)*fCtrs[facei] + (1.0/4.0)*cEst[nei[facei]];
solveVector pc = (3.0/4.0)*fc + (1.0/4.0)*cEst[nei[facei]];
// Accumulate volume-weighted face-pyramid centre
cellCtrs[nei[facei]] += pyr3Vol*pc;

View File

@ -61,22 +61,22 @@ bool Foam::primitiveMesh::checkClosedBoundary
// Loop through all boundary faces and sum up the face area vectors.
// For a closed boundary, this should be zero in all vector components
vector sumClosed(Zero);
scalar sumMagClosedBoundary = 0;
Vector<solveScalar> sumClosed(Zero);
solveScalar sumMagClosedBoundary = 0;
for (label facei = nInternalFaces(); facei < areas.size(); facei++)
{
if (!internalOrCoupledFaces.size() || !internalOrCoupledFaces[facei])
{
sumClosed += areas[facei];
sumClosed += Vector<solveScalar>(areas[facei]);
sumMagClosedBoundary += mag(areas[facei]);
}
}
reduce(sumClosed, sumOp<vector>());
reduce(sumMagClosedBoundary, sumOp<scalar>());
reduce(sumClosed, sumOp<Vector<solveScalar>>());
reduce(sumMagClosedBoundary, sumOp<solveScalar>());
vector openness = sumClosed/(sumMagClosedBoundary + VSMALL);
Vector<solveScalar> openness = sumClosed/(sumMagClosedBoundary + VSMALL);
if (cmptMax(cmptMag(openness)) > closedThreshold_)
{

View File

@ -6,7 +6,7 @@
\\/ M anipulation |
-------------------------------------------------------------------------------
Copyright (C) 2012-2016 OpenFOAM Foundation
Copyright (C) 2017-2018 OpenCFD Ltd.
Copyright (C) 2017-2019 OpenCFD Ltd.
-------------------------------------------------------------------------------
License
This file is part of OpenFOAM.
@ -420,6 +420,7 @@ Foam::tmp<Foam::scalarField> Foam::primitiveMeshTools::faceFlatness
tmp<scalarField> tfaceFlatness(new scalarField(mesh.nFaces(), 1.0));
scalarField& faceFlatness = tfaceFlatness.ref();
typedef Vector<solveScalar> solveVector;
forAll(fcs, facei)
{
@ -427,20 +428,20 @@ Foam::tmp<Foam::scalarField> Foam::primitiveMeshTools::faceFlatness
if (f.size() > 3 && magAreas[facei] > ROOTVSMALL)
{
const point& fc = fCtrs[facei];
const solveVector fc = fCtrs[facei];
// Calculate the sum of magnitude of areas and compare to magnitude
// of sum of areas.
scalar sumA = 0.0;
solveScalar sumA = 0.0;
forAll(f, fp)
{
const point& thisPoint = p[f[fp]];
const point& nextPoint = p[f.nextLabel(fp)];
const solveVector thisPoint = p[f[fp]];
const solveVector nextPoint = p[f.nextLabel(fp)];
// Triangle around fc.
vector n = 0.5*((nextPoint - thisPoint)^(fc - thisPoint));
solveVector n = 0.5*((nextPoint - thisPoint)^(fc - thisPoint));
sumA += mag(n);
}

View File

@ -84,7 +84,7 @@ void Foam::primitiveMesh::makeFaceCentresAndAreas
forAll(fs, facei)
{
const labelList& f = fs[facei];
label nPoints = f.size();
const label nPoints = f.size();
// If the face is a triangle, do a direct calculation for efficiency
// and to avoid round-off error-related problems
@ -95,26 +95,29 @@ void Foam::primitiveMesh::makeFaceCentresAndAreas
}
else
{
vector sumN = Zero;
scalar sumA = 0.0;
vector sumAc = Zero;
typedef Vector<solveScalar> solveVector;
point fCentre = p[f[0]];
solveVector sumN = Zero;
solveScalar sumA = 0.0;
solveVector sumAc = Zero;
solveVector fCentre = p[f[0]];
for (label pi = 1; pi < nPoints; pi++)
{
fCentre += p[f[pi]];
fCentre += solveVector(p[f[pi]]);
}
fCentre /= nPoints;
for (label pi = 0; pi < nPoints; pi++)
{
const point& nextPoint = p[f[(pi + 1) % nPoints]];
vector c = p[f[pi]] + nextPoint + fCentre;
vector n = (nextPoint - p[f[pi]])^(fCentre - p[f[pi]]);
scalar a = mag(n);
const label nextPi(pi == nPoints-1 ? 0 : pi+1);
const solveVector nextPoint(p[f[nextPi]]);
const solveVector thisPoint(p[f[pi]]);
solveVector c = thisPoint + nextPoint + fCentre;
solveVector n = (nextPoint - thisPoint)^(fCentre - thisPoint);
solveScalar a = mag(n);
sumN += n;
sumA += a;
sumAc += a*c;

View File

@ -154,7 +154,7 @@ namespace Foam
#endif
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
// Additional transcendental functions
// Additional transcendental functions and specialisations
namespace Foam
{
@ -172,6 +172,15 @@ namespace Foam
//- Lower incomplete gamma function
scalar incGamma_P(const scalar a, const scalar x);
//- Type to use for extended precision
template<>
class typeOfSolve<scalar>
{
public:
typedef solveScalar type;
};
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //

View File

@ -137,6 +137,15 @@ struct is_contiguous_scalar<SphericalTensor<Cmpt>>
{};
template<class Cmpt>
class typeOfSolve<SphericalTensor<Cmpt>>
{
public:
typedef SphericalTensor<solveScalar> type;
};
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
} // End namespace Foam

View File

@ -176,6 +176,15 @@ public:
};
template<class Cmpt>
class typeOfSolve<SymmTensor<Cmpt>>
{
public:
typedef SymmTensor<solveScalar> type;
};
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
} // End namespace Foam

View File

@ -344,6 +344,15 @@ public:
};
template<class Cmpt>
class typeOfSolve<Tensor<Cmpt>>
{
public:
typedef Tensor<solveScalar> type;
};
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
} // End namespace Foam

View File

@ -167,6 +167,15 @@ public:
};
template<class Cmpt>
class typeOfSolve<Vector<Cmpt>>
{
public:
typedef Vector<solveScalar> type;
};
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
} // End namespace Foam

View File

@ -35,6 +35,7 @@ Description
#ifndef products_H
#define products_H
#include "direction.H"
#include "pTraits.H"
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
@ -72,6 +73,16 @@ class symmTypeOfRank
{};
//- The extended precision type (solveScalar for float)
template<class Type>
class typeOfSolve
{
public:
typedef Type type;
};
//- The magnitude type for given argument.
template<class arg1>
class typeOfMag