From 1500f6ae1b947993cdcf7270acca1750331c174d Mon Sep 17 00:00:00 2001
From: Thomas Witkowski <thomas.witkowski@gmx.de>
Date: Tue, 29 Apr 2008 09:37:05 +0000
Subject: [PATCH] * Dies und jenes * OpenMP-Versuche aus Vorkonditionierern und
Loesern geloescht
---
AMDiS/libtool | 6 +--
AMDiS/src/Assembler.cc | 32 +++++++++++-----
AMDiS/src/BasisFunction.h | 9 +++++
AMDiS/src/DOFMatrix.cc | 12 +++---
AMDiS/src/DOFVector.hh | 76 +++++++++++---------------------------
AMDiS/src/Estimator.cc | 8 +---
AMDiS/src/Lagrange.cc | 17 +++++++--
AMDiS/src/Lagrange.h | 5 +++
AMDiS/src/Preconditioner.h | 18 ++-------
AMDiS/src/ProblemVec.cc | 43 +++++++--------------
AMDiS/src/SystemVector.h | 28 +++++---------
11 files changed, 110 insertions(+), 144 deletions(-)
diff --git a/AMDiS/libtool b/AMDiS/libtool
index 315f21f9..db01bc59 100755
--- a/AMDiS/libtool
+++ b/AMDiS/libtool
@@ -331,7 +331,7 @@ link_all_deplibs=unknown
sys_lib_search_path_spec=`echo " /usr/lib/gcc/i386-redhat-linux/4.1.2/ /usr/lib/gcc/i386-redhat-linux/4.1.2/ /usr/lib/gcc/i386-redhat-linux/4.1.2/../../../../i386-redhat-linux/lib/i386-redhat-linux/4.1.2/ /usr/lib/gcc/i386-redhat-linux/4.1.2/../../../../i386-redhat-linux/lib/ /usr/lib/gcc/i386-redhat-linux/4.1.2/../../../i386-redhat-linux/4.1.2/ /usr/lib/gcc/i386-redhat-linux/4.1.2/../../../ /lib/i386-redhat-linux/4.1.2/ /lib/ /usr/lib/i386-redhat-linux/4.1.2/ /usr/lib/" | $SED -e "s@${gcc_dir}@\${gcc_dir}@g;s@${gcc_ver}@\${gcc_ver}@g"`
# Run-time system search path for libraries
-sys_lib_dlsearch_path_spec="/lib /usr/lib /usr/lib/qt-3.3/lib "
+sys_lib_dlsearch_path_spec="/lib /usr/lib /usr/lib/octave-2.9.9 /usr/lib/qt-3.3/lib "
# Fix the shell variable $srcfile for the compiler.
fix_srcfile_path=""
@@ -7550,7 +7550,7 @@ link_all_deplibs=unknown
sys_lib_search_path_spec=`echo " /usr/lib/gcc/i386-redhat-linux/4.1.2/ /usr/lib/gcc/i386-redhat-linux/4.1.2/ /usr/lib/gcc/i386-redhat-linux/4.1.2/../../../../i386-redhat-linux/lib/i386-redhat-linux/4.1.2/ /usr/lib/gcc/i386-redhat-linux/4.1.2/../../../../i386-redhat-linux/lib/ /usr/lib/gcc/i386-redhat-linux/4.1.2/../../../i386-redhat-linux/4.1.2/ /usr/lib/gcc/i386-redhat-linux/4.1.2/../../../ /lib/i386-redhat-linux/4.1.2/ /lib/ /usr/lib/i386-redhat-linux/4.1.2/ /usr/lib/" | $SED -e "s@${gcc_dir}@\${gcc_dir}@g;s@${gcc_ver}@\${gcc_ver}@g"`
# Run-time system search path for libraries
-sys_lib_dlsearch_path_spec="/lib /usr/lib /usr/lib/qt-3.3/lib "
+sys_lib_dlsearch_path_spec="/lib /usr/lib /usr/lib/octave-2.9.9 /usr/lib/qt-3.3/lib "
# Fix the shell variable $srcfile for the compiler.
fix_srcfile_path=""
@@ -7861,7 +7861,7 @@ link_all_deplibs=unknown
sys_lib_search_path_spec=`echo " /usr/lib/gcc/i386-redhat-linux/3.4.6/ /usr/lib/gcc/i386-redhat-linux/3.4.6/ /usr/lib/gcc/i386-redhat-linux/3.4.6/../../../../i386-redhat-linux/lib/i386-redhat-linux/3.4.6/ /usr/lib/gcc/i386-redhat-linux/3.4.6/../../../../i386-redhat-linux/lib/ /usr/lib/gcc/i386-redhat-linux/3.4.6/../../../i386-redhat-linux/3.4.6/ /usr/lib/gcc/i386-redhat-linux/3.4.6/../../../ /lib/i386-redhat-linux/3.4.6/ /lib/ /usr/lib/i386-redhat-linux/3.4.6/ /usr/lib/" | $SED -e "s@${gcc_dir}@\${gcc_dir}@g;s@${gcc_ver}@\${gcc_ver}@g"`
# Run-time system search path for libraries
-sys_lib_dlsearch_path_spec="/lib /usr/lib /usr/lib/qt-3.3/lib "
+sys_lib_dlsearch_path_spec="/lib /usr/lib /usr/lib/octave-2.9.9 /usr/lib/qt-3.3/lib "
# Fix the shell variable $srcfile for the compiler.
fix_srcfile_path=""
diff --git a/AMDiS/src/Assembler.cc b/AMDiS/src/Assembler.cc
index 74dbfa1f..373fa100 100644
--- a/AMDiS/src/Assembler.cc
+++ b/AMDiS/src/Assembler.cc
@@ -497,6 +497,7 @@ namespace AMDiS {
int numPoints = quadrature->getNumPoints();
double *c = GET_MEMORY(double, numPoints);
+
for (int iq = 0; iq < numPoints; iq++) {
c[iq] = 0.0;
}
@@ -525,7 +526,7 @@ namespace AMDiS {
}
}
}
- } else { /* non symmetric assembling */
+ } else { // non symmetric assembling
for (int iq = 0; iq < numPoints; iq++) {
c[iq] *= elInfo->getDet();
@@ -542,6 +543,7 @@ namespace AMDiS {
}
}
}
+
FREE_MEMORY(c, double, numPoints);
FREE_MEMORY(phival, double, nCol);
}
@@ -1394,10 +1396,11 @@ namespace AMDiS {
Element *el = elInfo->getElement();
+
checkForNewTraverse();
checkQuadratures();
-
+
if ((el != lastMatEl && el != lastVecEl) || !operat->isOptimized()) {
initElement(elInfo);
}
@@ -1410,8 +1413,6 @@ namespace AMDiS {
} else {
if (rememberElMat) {
axpy(factor, *elementMatrix, *userMat);
- //*userMat += *elementMatrix * factor;
- //operat->addElementMatrix(elementMatrix, userMat, factor);
return;
}
}
@@ -1427,11 +1428,9 @@ namespace AMDiS {
if (zeroOrderAssembler)
zeroOrderAssembler->calculateElementMatrix(elInfo, mat);
- if(rememberElMat && userMat) {
+ if (rememberElMat && userMat) {
axpy(factor, *elementMatrix, *userMat);
- //*userMat += *elementMatrix * factor;
- //operat->addElementMatrix(elementMatrix, userMat, factor);
- }
+ }
}
void Assembler::calculateElementVector(const ElInfo *elInfo,
@@ -1597,27 +1596,40 @@ namespace AMDiS {
ElementMatrix *Assembler::initElementMatrix(ElementMatrix *elMat,
const ElInfo *elInfo)
{
- if(!elMat) {
+ if (!elMat) {
elMat = NEW ElementMatrix(nRow, nCol);
}
elMat->set(0.0);
-
+
DOFAdmin *rowAdmin = rowFESpace->getAdmin();
DOFAdmin *colAdmin = colFESpace->getAdmin();
+
Element *element = elInfo->getElement();
+ /*
elMat->rowIndices =
rowFESpace->getBasisFcts()->getLocalIndices(element,
rowAdmin,
NULL);
+ */
+ /*
elMat->colIndices =
colFESpace->getBasisFcts()->getLocalIndices(element,
colAdmin,
NULL);
+ */
+
+
+ rowFESpace->getBasisFcts()->getLocalIndicesVec(element,
+ rowAdmin,
+ &(elMat->rowIndices));
+ colFESpace->getBasisFcts()->getLocalIndicesVec(element,
+ colAdmin,
+ &(elMat->colIndices));
return elMat;
}
diff --git a/AMDiS/src/BasisFunction.h b/AMDiS/src/BasisFunction.h
index b9393a11..a1bf5929 100644
--- a/AMDiS/src/BasisFunction.h
+++ b/AMDiS/src/BasisFunction.h
@@ -29,6 +29,7 @@
#include <string>
#include "Global.h"
#include "Boundary.h"
+#include "MatrixVector.h"
namespace AMDiS {
@@ -318,6 +319,14 @@ namespace AMDiS {
return NULL;
};
+ /** \brief
+ * Returns local dof indices of the element for the given fe space.
+ */
+ virtual void getLocalIndicesVec(const Element*,
+ const DOFAdmin*,
+ Vector<DegreeOfFreedom>*) const
+ {};
+
/** \brief
* Evaluates elements value at barycentric coordinates lambda with local
diff --git a/AMDiS/src/DOFMatrix.cc b/AMDiS/src/DOFMatrix.cc
index 9a8beef2..099fecc2 100644
--- a/AMDiS/src/DOFMatrix.cc
+++ b/AMDiS/src/DOFMatrix.cc
@@ -450,15 +450,14 @@ namespace AMDiS {
FUNCNAME("DOFMatrix::assemble()");
if (!op && operators.size() == 0) {
- //WARNING("no operator\n");
return NULL;
}
-
+
Operator *operat = op ? op : operators[0];
-
+
elementMatrix =
operat->getAssembler()->initElementMatrix(elementMatrix, elInfo);
-
+
if (op) {
op->getElementMatrix(elInfo, elementMatrix);
} else {
@@ -472,10 +471,11 @@ namespace AMDiS {
elementMatrix,
*factorIt ? **factorIt : 1.0);
}
+
}
-
+
addElementMatrix(factor, *elementMatrix, bound);
-
+
return elementMatrix;
}
diff --git a/AMDiS/src/DOFVector.hh b/AMDiS/src/DOFVector.hh
index 4cc8e241..78624c2b 100644
--- a/AMDiS/src/DOFVector.hh
+++ b/AMDiS/src/DOFVector.hh
@@ -1207,61 +1207,29 @@ namespace AMDiS {
("a.size = %d too small: admin->sizeUsed = %d\n",
a.getSize(), rowAdmin->getUsedSize());
- // This is the old implementation of the mv-multiplication. It have been changed
- // because of the OpenMP-parallelization:
- // typename DOFVector<T>::Iterator vecIterator(dynamic_cast<DOFIndexed<T>*>(&result), USED_DOFS);
- // DOFMatrix::Iterator rowIterator(const_cast<DOFMatrix*>(&a), USED_DOFS);
- // for(vecIterator.reset(), rowIterator.reset();
- // !rowIterator.end();
- // ++rowIterator, ++vecIterator) {
- // sum = 0;
- // if(!add) *vecIterator = 0.0;
- // for(::std::vector<MatEntry>::iterator colIterator = rowIterator->begin();
- // colIterator != rowIterator->end();
- // colIterator++) {
- // jcol = colIterator->col;
- // if (jcol >= 0) { // entry used?
- // sum += (static_cast<double>(colIterator->entry)) * x[jcol];
- // } else {
- // if (jcol == DOFMatrix::NO_MORE_ENTRIES)
- // break;
- // }
- // }
- // *vecIterator += sum;
- // }
- int i;
- int maxI = result.getSize();
-
-#ifdef _OPENMP
-#pragma omp parallel for schedule(dynamic, 25000) default(shared) private(i)
-#endif
- for (i = 0; i < maxI; i++) {
- if (rowAdmin->isDOFFree(i)) {
- continue;
- }
-
- T sum = 0;
-
- const ::std::vector<MatEntry> *v = &a[i];
-
- for (int j = 0; j < static_cast<int>(v->size()); j++) {
- const MatEntry *m = &((*v)[j]);
- int jcol = m->col;
-
- if (jcol >= 0) { // entry used?
- sum += (static_cast<double>(m->entry)) * x[jcol];
- } else {
- if (jcol == DOFMatrix::NO_MORE_ENTRIES)
- break;
- }
- }
-
- if (add) {
- result[i] += sum;
- } else {
- result[i] = sum;
- }
+ typename DOFVector<T>::Iterator vecIterator(dynamic_cast<DOFIndexed<T>*>(&result), USED_DOFS);
+ DOFMatrix::Iterator rowIterator(const_cast<DOFMatrix*>(&a), USED_DOFS);
+ for(vecIterator.reset(), rowIterator.reset();
+ !rowIterator.end();
+ ++rowIterator, ++vecIterator) {
+
+ double sum = 0;
+ if (!add)
+ *vecIterator = 0.0;
+ for(::std::vector<MatEntry>::iterator colIterator = rowIterator->begin();
+ colIterator != rowIterator->end();
+ colIterator++) {
+
+ int jcol = colIterator->col;
+ if (jcol >= 0) { // entry used?
+ sum += (static_cast<double>(colIterator->entry)) * x[jcol];
+ } else {
+ if (jcol == DOFMatrix::NO_MORE_ENTRIES)
+ break;
+ }
+ }
+ *vecIterator += sum;
}
} else if (transpose == Transpose) {
diff --git a/AMDiS/src/Estimator.cc b/AMDiS/src/Estimator.cc
index b1ce2082..3b70fe17 100755
--- a/AMDiS/src/Estimator.cc
+++ b/AMDiS/src/Estimator.cc
@@ -235,10 +235,6 @@ namespace AMDiS {
el = elInfo->getElement();
- // if (el->getIndex() == 66065) {
- // ::std::cout << "UNSER ELEMENT!" << ::std::endl;
- // }
-
double det = elInfo->getDet();
const DimVec<WorldVector<double> > &Lambda = elInfo->getGrdLambda();
@@ -246,11 +242,11 @@ namespace AMDiS {
double h2 = h2_from_det(det, dim);
- for(iq = 0; iq < numPoints; iq++) {
+ for (iq = 0; iq < numPoints; iq++) {
riq[iq] = 0.0;
}
- for(system = 0; system < numSystems; system++) {
+ for (system = 0; system < numSystems; system++) {
if(matrix[system] == NULL) continue;
diff --git a/AMDiS/src/Lagrange.cc b/AMDiS/src/Lagrange.cc
index 886d615e..3df2274b 100644
--- a/AMDiS/src/Lagrange.cc
+++ b/AMDiS/src/Lagrange.cc
@@ -1004,7 +1004,7 @@ namespace AMDiS {
GeoIndex posIndex;
DegreeOfFreedom* result;
-
+
if (indices) {
result = indices;
} else {
@@ -1014,7 +1014,7 @@ namespace AMDiS {
localVecSize = nBasFcts;
result = localVec;
}
-
+
for (pos = 0, j = 0; pos <= dim; pos++) {
posIndex = INDEX_OF_DIM(pos, dim);
n0 = admin->getNumberOfPreDOFs(posIndex);
@@ -1031,10 +1031,21 @@ namespace AMDiS {
}
}
}
-
+
return result;
}
+ void Lagrange::getLocalIndicesVec(const Element* el,
+ const DOFAdmin *admin,
+ Vector<DegreeOfFreedom> *indices) const
+ {
+ if (indices->getSize() < nBasFcts) {
+ indices->resize(nBasFcts);
+ }
+
+ getLocalIndices(el, admin, &((*indices)[0]));
+ }
+
void Lagrange::l2ScpFctBas(Quadrature *q,
AbstractFunction<WorldVector<double>, WorldVector<double> >* f,
DOFVector<WorldVector<double> >* fh)
diff --git a/AMDiS/src/Lagrange.h b/AMDiS/src/Lagrange.h
index 51a38f00..78cb7f28 100644
--- a/AMDiS/src/Lagrange.h
+++ b/AMDiS/src/Lagrange.h
@@ -152,6 +152,11 @@ namespace AMDiS {
const DOFAdmin*,
DegreeOfFreedom*) const;
+
+ void getLocalIndicesVec(const Element*,
+ const DOFAdmin*,
+ Vector<DegreeOfFreedom>*) const;
+
/** \brief
* Implements BasisFunction::l2ScpFctBas
*/
diff --git a/AMDiS/src/Preconditioner.h b/AMDiS/src/Preconditioner.h
index 5ddec677..eb3c0f96 100644
--- a/AMDiS/src/Preconditioner.h
+++ b/AMDiS/src/Preconditioner.h
@@ -237,11 +237,7 @@ namespace AMDiS {
* Initialisation of the preconditioner
*/
virtual void init() {
- int i;
-#ifdef _OPENMP
-#pragma omp parallel for schedule(static, 1) default(shared) private(i)
-#endif
- for (i = 0; i < scalPrecons.getSize(); i++) {
+ for (int i = 0; i < scalPrecons.getSize(); i++) {
scalPrecons[i]->init();
}
};
@@ -250,11 +246,7 @@ namespace AMDiS {
* Preconditioning method
*/
virtual void precon(SystemVector *x) {
- int i;
-#ifdef _OPENMP
-#pragma omp parallel for schedule(static, 1) default(shared) private(i)
-#endif
- for (i = 0; i < scalPrecons.getSize(); i++) {
+ for (int i = 0; i < scalPrecons.getSize(); i++) {
scalPrecons[i]->precon(x->getDOFVector(i));
}
};
@@ -263,11 +255,7 @@ namespace AMDiS {
* Frees needed memory.
*/
virtual void exit() {
- int i;
-#ifdef _OPENMP
-#pragma omp parallel for schedule(static, 1) default(shared) private(i)
-#endif
- for (i = 0; i < scalPrecons.getSize(); i++) {
+ for (int i = 0; i < scalPrecons.getSize(); i++) {
scalPrecons[i]->exit();
}
};
diff --git a/AMDiS/src/ProblemVec.cc b/AMDiS/src/ProblemVec.cc
index b965e1c9..bbd98a4f 100644
--- a/AMDiS/src/ProblemVec.cc
+++ b/AMDiS/src/ProblemVec.cc
@@ -679,10 +679,6 @@ namespace AMDiS {
}
}
- TraverseStack stack;
- DualTraverse dualStack;
- ElInfo *elInfo;
-
for (int i = 0; i < numComponents_; i++) {
for (int j = 0; j < numComponents_; j++) {
// Only if this variable is true, the current matrix will be assembled.
@@ -708,19 +704,22 @@ namespace AMDiS {
continue;
}
-
+
if (componentMeshes_[i] != componentMeshes_[j]) {
ERROR_EXIT("not yet\n");
} else {
if (assembleMatrix && matrix->getBoundaryManager())
matrix->getBoundaryManager()->initMatrix(matrix);
- elInfo = stack.traverseFirst(componentMeshes_[i], -1, assembleFlag);
+ TraverseStack stack;
+ ElInfo *elInfo = stack.traverseFirst(componentMeshes_[i], -1, assembleFlag);
+
while (elInfo) {
const BoundaryType *bound =
useGetBound_ ?
componentSpaces_[i]->getBasisFcts()->getBound(elInfo, NULL) :
NULL;
+
if (assembleMatrix) {
matrix->assemble(1.0, elInfo, bound);
if (matrix->getBoundaryManager()) {
@@ -732,12 +731,14 @@ namespace AMDiS {
if (i == j) {
rhs_->getDOFVector(i)->assemble(1.0, elInfo, bound);
}
+
elInfo = stack.traverseNext(elInfo);
}
-
+
if (assembleMatrix && matrix->getBoundaryManager())
- matrix->getBoundaryManager()->exitMatrix(matrix);
+ matrix->getBoundaryManager()->exitMatrix(matrix);
}
+
assembledMatrix_[i][j] = true;
}
@@ -748,7 +749,8 @@ namespace AMDiS {
if (solution_->getDOFVector(i)->getBoundaryManager())
solution_->getDOFVector(i)->getBoundaryManager()->initVector(solution_->getDOFVector(i));
- elInfo = stack.traverseFirst(componentMeshes_[i], -1, assembleFlag);
+ TraverseStack stack;
+ ElInfo *elInfo = stack.traverseFirst(componentMeshes_[i], -1, assembleFlag);
while (elInfo) {
if(rhs_->getDOFVector(i)->getBoundaryManager())
rhs_->getDOFVector(i)->getBoundaryManager()->
@@ -761,28 +763,11 @@ namespace AMDiS {
if (rhs_->getDOFVector(i)->getBoundaryManager())
rhs_->getDOFVector(i)->getBoundaryManager()->exitVector(rhs_->getDOFVector(i));
if (solution_->getDOFVector(i)->getBoundaryManager())
- solution_->getDOFVector(i)->getBoundaryManager()->exitVector(solution_->getDOFVector(i));
+ solution_->getDOFVector(i)->getBoundaryManager()->exitVector(solution_->getDOFVector(i));
}
-
- for (int i = 0; i < numComponents_; i++) {
- // ::std::cout << "Vector: " << i << ::std::endl;
- // rhs_->getDOFVector(i)->print();
- }
-
+
INFO(info_, 8)("buildAfterCoarsen needed %.5f seconds\n",
- TIME_USED(first,clock()));
-
-
-// int memsize = 0;
-// for (int i = 0; i < numComponents_; i++) {
-// for (int j = 0; j < numComponents_; j++) {
-// DOFMatrix *matrix = (*systemMatrix_)[i][j];
-// memsize += (*matrix).memsize();
-// }
-// }
-
-// ::std::cout << "SIZE: " << memsize << ::std::endl;
-
+ TIME_USED(first, clock()));
}
void ProblemVec::writeFiles(AdaptInfo *adaptInfo, bool force)
diff --git a/AMDiS/src/SystemVector.h b/AMDiS/src/SystemVector.h
index acca3e7b..8c532cb3 100644
--- a/AMDiS/src/SystemVector.h
+++ b/AMDiS/src/SystemVector.h
@@ -474,14 +474,11 @@ namespace AMDiS {
TEST_EXIT(size == matrix.getNumRows())("incompatible sizes\n");
TEST_EXIT(size == matrix.getNumCols())("incompatible sizes\n");
- int i, j;
-
-#ifdef _OPENMP
-#pragma omp parallel for schedule(static, 1) default(shared) private(i,j)
-#endif
- for (i = 0; i < size; i++) {
- if (!add) result.getDOFVector(i)->set(0.0);
- for (j = 0; j < size; j++) {
+ for (int i = 0; i < size; i++) {
+ if (!add)
+ result.getDOFVector(i)->set(0.0);
+
+ for (int j = 0; j < size; j++) {
if (matrix[i][j]) {
mv<double>(NoTranspose,
*(matrix[i][j]),
@@ -500,12 +497,10 @@ namespace AMDiS {
{
TEST_EXIT(x.getNumVectors() == y.getNumVectors())
("invalid size\n");
- int i, size = x.getNumVectors();
-#ifdef _OPENMP
-#pragma omp parallel for schedule(static, 1) default(shared) private(i)
-#endif
- for (i = 0; i < size; i++) {
+ int size = x.getNumVectors();
+
+ for (int i = 0; i < size; i++) {
axpy(a, *(x.getDOFVector(i)), *(y.getDOFVector(i)));
}
};
@@ -517,12 +512,9 @@ namespace AMDiS {
{
TEST_EXIT(x.getNumVectors() == y.getNumVectors())
("invalid size\n");
- int i, size = x.getNumVectors();
+ int size = x.getNumVectors();
-#ifdef _OPENMP
-#pragma omp parallel for schedule(static, 1) default(shared) private(i)
-#endif
- for (i = 0; i < size; i++) {
+ for (int i = 0; i < size; i++) {
xpay(a, *(x.getDOFVector(i)), *(y.getDOFVector(i)));
}
}
--
GitLab