Skip to content

I
INMOST
Commit 519bb261 authored by Alexander Danilov's avatar Alexander Danilov
Browse files
Options

Update examples and readme files 

Install instructions from readme file were tested on Linux machine.
parent 86400a9a
Hide whitespace changes
Inline Side-by-side
Showing with 63 additions and 61 deletions
README
View file @ 519bb261
......@@ -31,7 +31,7 @@ rm -f petsc-3.4.5.tar.gz
cd petsc-3.4.5
export PETSC_DIR="`pwd`"
export PETSC_ARCH=linux-gnu-opt
./configure --download-f-blas-lapack --download-metis --download-parmetis --useThreads=0 --with-debugging=0 --with-mpi-dir=/usr --with-x=0 -with-shared-libraries=0
./configure --download-f-blas-lapack --download-metis --download-parmetis --useThreads=0 --with-debugging=0 --with-mpi-dir=/usr --with-x=0 --with-shared-libraries=0
make all
......@@ -42,16 +42,15 @@ Download and unpack INMOST source archive.
cd "$INMOST_LIBS"
wget https://github.com/INM-RAS/INMOST/archive/master.tar.gz
tar zxf INMOST-master.tar.gz
tar zxf master.tar.gz
rm -f INMOST-master.tar.gz
We will create separate directory for INMOST compilation.
Depending on your version of gcc compiler you may need one of these flags for CMAKE_CXX_FLAGS variable: "-std=c++11" or "-std=c++0x".
mkdir -p INMOST-build
cd INMOST-build
cmake -DUSE_AUTODIFF=OFF -DUSE_SOLVER_PETSC=ON -DUSE_PARTITIONER_PARMETIS=ON -DCMAKE_CXX_FLAGS="-std=c++11" -DCOMPILE_EXAMPLES=ON ../INMOST-master
cmake -DUSE_AUTODIFF=OFF -DUSE_SOLVER_PETSC=ON -DUSE_PARTITIONER=OFF -DCOMPILE_EXAMPLES=ON ../INMOST-master
make
......@@ -65,11 +64,11 @@ Each example may be executed in serial or parallel ways.
Parallel Grid Generation
------------------------
This example creates simple cubic or prismatic mesh. You can use ParaView to
This example creates simple cubic or prismatic mesh. You can use ParaView to view the meshes.
cd "$INMOST_LIBS/INMOST-build"
cd examples/GridGen
mpirun -np 4 GrdiGen 3 32 32 4
mpirun -np 4 ./GridGen 4 32 32 32
Generator parameters are: ng nx ny nz
where ng=3 stands for Prismatic generator and
......@@ -89,11 +88,11 @@ Parallel Finite Volume Discretization
This example uses simple two-point FVM scheme to solve Laplace's equation in unit cube domain.
cd ../FVDiscr
mpirun -np 4 FVDiscr ../GridGen/grid.pvtk A.mtx b.rhs
mpirun -np 4 ./FVDiscr ../GridGen/grid.pvtk A.mtx b.rhs
Files result.pvtk (as well as result_X.vtk with X=0,1,2,3) and A.mtx b.rhs will appear in the current directory.
Run
paraview --data=grid.pvtk
paraview --data=result.pvtk
and try the following tags in objects to display:
Solution - the solution to the problem
K - tensor K (constant equal to 1 in this example)
......@@ -103,8 +102,8 @@ Solve the Matrix stored in mtx format
This example solves the linear system using different solvers.
cd ../FVDiscr
mpirun -np 4 MatSolve 0 ../FVDiscr/A.mtx ../FVDiscr/b.rhs
cd ../MatSolve
mpirun -np 4 ./MatSolve 0 ../FVDiscr/A.mtx ../FVDiscr/b.rhs
Solution time and the true residual will output to the screen.
The first parameter selects the solver:
......
README.md
View file @ 519bb261
......@@ -29,7 +29,7 @@ rm -f petsc-3.4.5.tar.gz
cd petsc-3.4.5
export PETSC_DIR="`pwd`"
export PETSC_ARCH=linux-gnu-opt
./configure --download-f-blas-lapack --download-metis --download-parmetis --useThreads=0 --with-debugging=0 --with-mpi-dir=/usr --with-x=0 –with-shared-libraries=0
./configure --download-f-blas-lapack --download-metis --download-parmetis --useThreads=0 --with-debugging=0 --with-mpi-dir=/usr --with-x=0 --with-shared-libraries=0
make all
```
......@@ -39,20 +39,19 @@ Download and unpack INMOST source archive.
```
cd "$INMOST_LIBS"
wget https://github.com/INM-RAS/INMOST/archive/master.tar.gz
tar zxf INMOST-master.tar.gz
tar zxf master.tar.gz
rm -f INMOST-master.tar.gz
```
We will create separate directory for INMOST compilation.
Depending on your version of gcc compiler you may need one of these flags for `CMAKE_CXX_FLAGS variable`: `"-std=c++11"` or `"-std=c++0x"`.
```
mkdir -p INMOST-build
cd INMOST-build
cmake -DUSE_AUTODIFF=OFF -DUSE_SOLVER_PETSC=ON -DUSE_PARTITIONER_PARMETIS=ON -DCMAKE_CXX_FLAGS="-std=c++11" -DCOMPILE_EXAMPLES=ON ../INMOST-master
cmake -DUSE_AUTODIFF=OFF -DUSE_SOLVER_PETSC=ON -DUSE_PARTITIONER=OFF -DCOMPILE_EXAMPLES=ON ../INMOST-master
make
```
## Examples
## Examples
Several representative examples are provided in source archive.
Here we will try three parallel steps: grid generation, FVM discretization and linear matrix solution.
......@@ -60,11 +59,11 @@ Each example may be executed in serial or parallel ways.
### Parallel Grid Generation
This example creates simple cubic or prismatic mesh. You can use ParaView to
This example creates simple cubic or prismatic mesh. You can use ParaView to view the meshes.
```
cd "$INMOST_LIBS/INMOST-build"
cd examples/GridGen
mpirun -np 4 GrdiGen 3 32 32 4
mpirun -np 4 ./GridGen 4 32 32 32
```
Generator parameters are: `ng nx ny nz`
where `ng=3` stands for Prismatic generator and
......@@ -83,11 +82,11 @@ and try the following tags in objects to display:
This example uses simple two-point FVM scheme to solve Laplace's equation in unit cube domain.
```
cd ../FVDiscr
mpirun -np 4 FVDiscr ../GridGen/grid.pvtk A.mtx b.rhs
mpirun -np 4 ./FVDiscr ../GridGen/grid.pvtk A.mtx b.rhs
```
Files result.pvtk (as well as result_X.vtk with X=0,1,2,3) and A.mtx b.rhs will appear in the current directory.
Run
`paraview --data=grid.pvtk`
`paraview --data=result.pvtk`
and try the following tags in objects to display:
- `Solution` – the solution to the problem
- `K` – tensor K (constant equal to 1 in this example)
......@@ -96,8 +95,8 @@ and try the following tags in objects to display:
This example solves the linear system using different solvers.
```
cd ../FVDiscr
mpirun -np 4 MatSolve 0 ../FVDiscr/A.mtx ../FVDiscr/b.rhs
cd ../MatSolve
mpirun -np 4 ./MatSolve 0 ../FVDiscr/A.mtx ../FVDiscr/b.rhs
```
Solution time and the true residual will output to the screen.
The first parameter selects the solver:
......
examples/CMakeLists.txt
View file @ 519bb261
#add_subdirectory(DrawGrid)
add_subdirectory(OldDrawGrid)
#add_subdirectory(DrawMatrix)
#add_subdirectory(MatSolve)
#add_subdirectory(GridGen)
#add_subdirectory(FVDiscr)
add_subdirectory(MatSolve)
add_subdirectory(GridGen)
add_subdirectory(FVDiscr)
#add_subdirectory(OctreeCutcell)
#add_subdirectory(Solver)
examples/FVDiscr/main.cpp
View file @ 519bb261
......@@ -128,7 +128,7 @@ int main(int argc,char ** argv)
Solver::Matrix A; // Declare the matrix of the linear system to be solved
Solver::Vector x,b; // Declare the solution and the right-hand side vectors
std::map<GeometricData,ElementType> table;
tiny_map<GeometricData,ElementType,5> table;
table[MEASURE] = CELL | FACE;
table[CENTROID] = CELL | FACE;
......@@ -160,10 +160,10 @@ int main(int argc,char ** argv)
{
//~ std::cout << face->LocalID() << " / " << m->NumberOfFaces() << std::endl;
Element::Status s1,s2;
Cell * r1 = face->BackCell();
Cell * r2 = face->FrontCell();
if( ((r1 == NULL || (s1 = r1->GetStatus()) == Element::Ghost)?0:1) +
((r2 == NULL || (s2 = r2->GetStatus()) == Element::Ghost)?0:1) == 0) continue;
Cell r1 = face->BackCell();
Cell r2 = face->FrontCell();
if( ((!r1->isValid() || (s1 = r1->GetStatus()) == Element::Ghost)?0:1) +
((!r2->isValid() || (s2 = r2->GetStatus()) == Element::Ghost)?0:1) == 0) continue;
Storage::real f_nrm[3], r1_cnt[3], r2_cnt[3], f_cnt[3], d1[3], Coef;
Storage::real f_area = face->Area(); // Get the face area
Storage::real vol1 = r1->Volume(), vol2; // Get the cell volume
......@@ -175,7 +175,7 @@ int main(int argc,char ** argv)
f_nrm[2] /= f_area;
r1->Barycenter(r1_cnt); // Get the barycenter of the cell
face->Barycenter(f_cnt); // Get the barycenter of the face
if( r2 == NULL ) // boundary condition
if( !r2->isValid() ) // boundary condition
{
Storage::real bnd_pnt[3], dist;
make_vec(f_cnt,r1_cnt,d1);
......@@ -266,7 +266,7 @@ int main(int argc,char ** argv)
if( m->GetProcessorRank() == 0 ) std::cout << "Retrive data: " << Timer()-ttt << std::endl;
ttt = Timer();
m->ExchangeData(phi,CELL); // Data exchange over processors
m->ExchangeData(phi,CELL,0); // Data exchange over processors
BARRIER
if( m->GetProcessorRank() == 0 ) std::cout << "Exchange phi: " << Timer()-ttt << std::endl;
......
examples/GridGen/main.cpp
View file @ 519bb261
......@@ -71,7 +71,7 @@ Mesh * ParallelCubeGenerator(INMOST_MPI_Comm comm, int nx, int ny, int nz)
localend[j] = localstart[j] + localsize[j];
}
std::vector<Node *> newverts;
ElementArray<Node> newverts(m);
newverts.reserve(localsize[0]*localsize[1]*localsize[2]);
for(int i = localstart[0]; i <= localend[0]; i++)
......@@ -83,7 +83,7 @@ Mesh * ParallelCubeGenerator(INMOST_MPI_Comm comm, int nx, int ny, int nz)
xyz[1] = j * 1.0 / (sizes[1]);
xyz[2] = k * 1.0 / (sizes[2]);
newverts.push_back(m->CreateNode(xyz)); // Create node in the mesh
if( ((int)newverts.size()-1) != V_ID(i,j,k))
if( newverts.size() != V_ID(i,j,k) + 1 )
printf("v_id = %ld, [%d,%d,%d] = %d\n", newverts.size()-1,i,j,k, V_ID(i, j, k));
}
......@@ -95,17 +95,17 @@ Mesh * ParallelCubeGenerator(INMOST_MPI_Comm comm, int nx, int ny, int nz)
for(int j = localstart[1]+1; j <= localend[1]; j++)
for(int k = localstart[2]+1; k <= localend[2]; k++)
{
const INMOST_DATA_ENUM_TYPE nvf[24] = {0,4,6,2,1,3,7,5,0,1,5,4,2,6,7,3,0,2,3,1,4,5,7,6};
const INMOST_DATA_ENUM_TYPE numnodes[6] = { 4, 4, 4, 4, 4, 4 };
Node * verts[8];
verts[0] = newverts[V_ID(i - 1,j - 1, k - 1)];
verts[1] = newverts[V_ID(i - 0,j - 1, k - 1)];
verts[2] = newverts[V_ID(i - 1,j - 0, k - 1)];
verts[3] = newverts[V_ID(i - 0,j - 0, k - 1)];
verts[4] = newverts[V_ID(i - 1,j - 1, k - 0)];
verts[5] = newverts[V_ID(i - 0,j - 1, k - 0)];
verts[6] = newverts[V_ID(i - 1,j - 0, k - 0)];
verts[7] = newverts[V_ID(i - 0,j - 0, k - 0)];
const INMOST_DATA_INTEGER_TYPE nvf[24] = {0,4,6,2,1,3,7,5,0,1,5,4,2,6,7,3,0,2,3,1,4,5,7,6};
const INMOST_DATA_INTEGER_TYPE numnodes[6] = { 4, 4, 4, 4, 4, 4 };
ElementArray<Node> verts(m);
verts.push_back(newverts[V_ID(i - 1,j - 1, k - 1)]);
verts.push_back(newverts[V_ID(i - 0,j - 1, k - 1)]);
verts.push_back(newverts[V_ID(i - 1,j - 0, k - 1)]);
verts.push_back(newverts[V_ID(i - 0,j - 0, k - 1)]);
verts.push_back(newverts[V_ID(i - 1,j - 1, k - 0)]);
verts.push_back(newverts[V_ID(i - 0,j - 1, k - 0)]);
verts.push_back(newverts[V_ID(i - 1,j - 0, k - 0)]);
verts.push_back(newverts[V_ID(i - 0,j - 0, k - 0)]);
m->CreateCell(verts,nvf,numnodes,6).first; // Create the cubic cell in the mesh
}
......@@ -176,7 +176,7 @@ Mesh * ParallelCubePrismGenerator(INMOST_MPI_Comm comm, int nx, int ny, int nz)
localend[j] = localstart[j] + localsize[j];
}
std::vector<Node *> newverts;
ElementArray<Node> newverts(m);
newverts.reserve(localsize[0]*localsize[1]*localsize[2]);
for(int i = localstart[0]; i <= localend[0]; i++)
......@@ -188,7 +188,7 @@ Mesh * ParallelCubePrismGenerator(INMOST_MPI_Comm comm, int nx, int ny, int nz)
xyz[1] = j * 1.0 / (sizes[1]);
xyz[2] = k * 1.0 / (sizes[2]);
newverts.push_back(m->CreateNode(xyz)); // Create node in the mesh
if( ((int)newverts.size()-1) != V_ID(i,j,k))
if( newverts.size() != V_ID(i,j,k) + 1 )
printf("v_id = %ld, [%d,%d,%d] = %d\n", newverts.size()-1,i,j,k, V_ID(i, j, k));
}
......@@ -202,23 +202,23 @@ Mesh * ParallelCubePrismGenerator(INMOST_MPI_Comm comm, int nx, int ny, int nz)
for(int j = localstart[1]+1; j <= localend[1]; j++)
for(int k = localstart[2]+1; k <= localend[2]; k++)
{
const INMOST_DATA_ENUM_TYPE NE_nvf1[18] = {0,4,6,2,0,3,7,4,2,6,7,3,0,2,3,4,7,6};
const INMOST_DATA_ENUM_TYPE NE_nvf2[18] = {0,4,7,3,1,3,7,5,0,1,5,4,0,3,1,4,5,7};
const INMOST_DATA_INTEGER_TYPE NE_nvf1[18] = {0,4,6,2,0,3,7,4,2,6,7,3,0,2,3,4,7,6};
const INMOST_DATA_INTEGER_TYPE NE_nvf2[18] = {0,4,7,3,1,3,7,5,0,1,5,4,0,3,1,4,5,7};
const INMOST_DATA_ENUM_TYPE NE_nvf3[18] = {0,4,6,2,2,6,5,1,1,5,4,0,0,2,1,4,5,6};
const INMOST_DATA_ENUM_TYPE NE_nvf4[18] = {1,5,6,2,1,3,7,5,7,3,2,6,1,2,3,6,5,7};
const INMOST_DATA_INTEGER_TYPE NE_nvf3[18] = {0,4,6,2,2,6,5,1,1,5,4,0,0,2,1,4,5,6};
const INMOST_DATA_INTEGER_TYPE NE_nvf4[18] = {1,5,6,2,1,3,7,5,7,3,2,6,1,2,3,6,5,7};
const INMOST_DATA_ENUM_TYPE numnodes[5] = {4,4,4,3,3};
const INMOST_DATA_INTEGER_TYPE numnodes[5] = {4,4,4,3,3};
Node * verts[8];
verts[0] = newverts[V_ID(i - 1,j - 1, k - 1)];
verts[1] = newverts[V_ID(i - 0,j - 1, k - 1)];
verts[2] = newverts[V_ID(i - 1,j - 0, k - 1)];
verts[3] = newverts[V_ID(i - 0,j - 0, k - 1)];
verts[4] = newverts[V_ID(i - 1,j - 1, k - 0)];
verts[5] = newverts[V_ID(i - 0,j - 1, k - 0)];
verts[6] = newverts[V_ID(i - 1,j - 0, k - 0)];
verts[7] = newverts[V_ID(i - 0,j - 0, k - 0)];
ElementArray<Node> verts(m);
verts.push_back(newverts[V_ID(i - 1,j - 1, k - 1)]);
verts.push_back(newverts[V_ID(i - 0,j - 1, k - 1)]);
verts.push_back(newverts[V_ID(i - 1,j - 0, k - 1)]);
verts.push_back(newverts[V_ID(i - 0,j - 0, k - 1)]);
verts.push_back(newverts[V_ID(i - 1,j - 1, k - 0)]);
verts.push_back(newverts[V_ID(i - 0,j - 1, k - 0)]);
verts.push_back(newverts[V_ID(i - 1,j - 0, k - 0)]);
verts.push_back(newverts[V_ID(i - 0,j - 0, k - 0)]);
// Create two prismatic cells in the mesh
if ((i + j) % 2 == 0)
......@@ -280,10 +280,14 @@ int main(int argc, char *argv[])
filename += ".vtk";
else
filename += ".pvtk";
#if defined(USE_MPI)
MPI_Barrier(mesh->GetCommunicator());
#endif
tt = Timer();
mesh->Save(filename); // Save constructed mesh to the file
#if defined(USE_MPI)
MPI_Barrier(mesh->GetCommunicator());
#endif
tt = Timer() - tt;
if( mesh->GetProcessorRank() == 0 ) std::cout << "Save to file \"" << filename << "\" time: " << tt << std::endl;
......
examples/MatSolve/main.cpp
View file @ 519bb261
......@@ -40,7 +40,7 @@ int main(int argc, char ** argv)
Solver::Vector b("rhs"); // Declare the right-hand side vector
Solver::Vector x("sol"); // Declare the solution vector
//std::cout << rank << " load matrix from " << std::string(argv[2]) << " ..." << std::endl;
long double t = Timer(), tt = Timer();
double t = Timer(), tt = Timer();
mat.Load(std::string(argv[2])); //if interval parameters not set, matrix will be divided automatically
BARRIER
if( !rank ) std::cout << "load matrix: " << Timer() - t << std::endl;
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment