Skip to content

Gpu dev #327

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
maorz1998 merged 3 commits into from
Aug 27, 2023
Merged

Gpu dev #327

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
03744a4
fix bugs in UEqn
maorz1998 Aug 22, 2023
5c7daa5
opt ldu2csr
maorz1998 Aug 22, 2023
920fc84
fvMatrix.A(), fvMatrix.H(), opt ldu2csr
maorz1998 Aug 27, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
42 changes: 23 additions & 19 deletions applications/solvers/dfLowMachFoam/UEqn.H
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -135,7 +135,7 @@

TICK_START;
U.oldTime();
memcpy(h_u, &U.oldTime()[0][0], dfDataBase.cell_value_vec_bytes);
memcpy(h_u, &U.oldTime()[0][0], dfDataBase.cell_value_vec_bytes);
memcpy(h_p, &p[0], dfDataBase.cell_value_bytes);
memcpy(h_nu_eff, &nuEff[0], dfDataBase.cell_value_bytes);
TICK_STOP(copy to pinned memory);
Expand Down Expand Up @@ -177,30 +177,34 @@
// postProcess
TICK_START;
UEqn_GPU.postProcess(h_u);
memcpy(&U[0][0], h_u, dfDataBase.cell_value_vec_bytes);
U.correctBoundaryConditions();
K = 0.5*magSqr(U);
DEBUG_TRACE;
TICK_STOP(post process time);

// checkResult
// TODO: for temp, now we compare ldu, finally we compare csr
std::vector<double> h_internal_coeffs(dfDataBase.num_boundary_surfaces * 3);
std::vector<double> h_boundary_coeffs(dfDataBase.num_boundary_surfaces * 3);
offset = 0;
for (int patchi = 0; patchi < dfDataBase.num_patches; patchi++)
{
int patchsize = dfDataBase.patch_size[patchi];
const double* internal_coeff_ptr = &UEqn.internalCoeffs()[patchi][0][0];
const double* boundary_coeff_ptr = &UEqn.boundaryCoeffs()[patchi][0][0];
memcpy(h_internal_coeffs.data() + offset * 3, internal_coeff_ptr, patchsize * 3 * sizeof(double));
memcpy(h_boundary_coeffs.data() + offset * 3, boundary_coeff_ptr, patchsize * 3 * sizeof(double));
offset += patchsize;
}
bool printFlag = false;
UEqn_GPU.compareResult(&UEqn.lower()[0], &UEqn.upper()[0], &UEqn.diag()[0], &UEqn.source()[0][0],
h_internal_coeffs.data(), h_boundary_coeffs.data(),
// &DivTensor[0][0],
printFlag);
DEBUG_TRACE;
// std::vector<double> h_internal_coeffs(dfDataBase.num_boundary_surfaces * 3);
// std::vector<double> h_boundary_coeffs(dfDataBase.num_boundary_surfaces * 3);
// Info << "location 0 " << endl;
// offset = 0;
// for (int patchi = 0; patchi < dfDataBase.num_patches; patchi++)
// {
// int patchsize = dfDataBase.patch_size[patchi];
// const double* internal_coeff_ptr = &UEqn.internalCoeffs()[patchi][0][0];
// const double* boundary_coeff_ptr = &UEqn.boundaryCoeffs()[patchi][0][0];
// memcpy(h_internal_coeffs.data() + offset * 3, internal_coeff_ptr, patchsize * 3 * sizeof(double));
// memcpy(h_boundary_coeffs.data() + offset * 3, boundary_coeff_ptr, patchsize * 3 * sizeof(double));
// offset += patchsize;
// }
// bool printFlag = false;
// UEqn_GPU.compareResult(&UEqn.lower()[0], &UEqn.upper()[0], &UEqn.diag()[0], &UEqn.source()[0][0],
// h_internal_coeffs.data(), h_boundary_coeffs.data(),
// // &DivTensor[0][0],
// printFlag);
// DEBUG_TRACE;

#else
start1 = std::clock();
tmp<fvVectorMatrix> tUEqn
Expand Down
116 changes: 0 additions & 116 deletions applications/solvers/dfLowMachFoam/new_UEqn.H
View file
Open in desktop

This file was deleted.

113 changes: 0 additions & 113 deletions applications/solvers/dfLowMachFoam/new_dfLowMachFoam.C
View file
Open in desktop

This file was deleted.

43 changes: 43 additions & 0 deletions applications/solvers/dfLowMachFoam/pEqn.H
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,11 +67,52 @@ const volScalarField psip0(psi*p);
time_monitor_UEqn_A_correctBC += double(end1 - start1) / double(CLOCKS_PER_SEC);
#endif

#ifdef GPUSolverNew_
volVectorField UEqn_H
(
IOobject
(
"H("+U.name()+')',
runTime.timeName(),
mesh,
IOobject::NO_READ,
IOobject::NO_WRITE
),
mesh,
dimensionedVector(dimensionSet(1,-2,-2,0,0,0,0), Zero),
extrapolatedCalculatedFvPatchScalarField::typeName
);
UEqn_GPU.H(&UEqn_H[0][0]);
UEqn_H.correctBoundaryConditions();

volScalarField UEqn_A
(
IOobject
(
"A("+U.name()+')',
runTime.timeName(),
mesh,
IOobject::NO_READ,
IOobject::NO_WRITE
),
mesh,
dimensionedScalar(dimensionSet(1,-3,-1,0,0,0,0), Zero),
extrapolatedCalculatedFvPatchScalarField::typeName
);
UEqn_GPU.A(&UEqn_A[0]);
UEqn_A.correctBoundaryConditions();
#endif


start2 = std::clock();
#ifdef GPUSolver_
volScalarField rAU(1.0/UEqn_A);
surfaceScalarField rhorAUf("rhorAUf", fvc::interpolate(rho*rAU));
volVectorField HbyA(constrainHbyA(rAU*UEqn_H, U, p));
#elif defined GPUSolverNew_
volScalarField rAU(1.0/UEqn_A);
surfaceScalarField rhorAUf("rhorAUf", fvc::interpolate(rho*rAU));
volVectorField HbyA(constrainHbyA(rAU*UEqn_H, U, p));
#else
volScalarField rAU(1.0/UEqn.A());
surfaceScalarField rhorAUf("rhorAUf", fvc::interpolate(rho*rAU));
Expand Down Expand Up @@ -179,6 +220,8 @@ K = 0.5*magSqr(U);
end1 = std::clock();
time_monitor_UEqn += double(end1 - start1) / double(CLOCKS_PER_SEC);
time_monitor_UEqn_correctBC += double(end1 - start1) / double(CLOCKS_PER_SEC);
#elif defined GPUSolverNew_
UEqn_GPU.correctPsi(&U[0][0]);
#endif

if (pimple.simpleRho())
Expand Down
1 change: 1 addition & 0 deletions src_gpu/dfMatrixDataBase.H
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -103,6 +103,7 @@ struct dfMatrixDataBase
int *d_lower_to_csr_index = nullptr;
int *d_diag_to_csr_index= nullptr;
int *d_upper_to_csr_index= nullptr;
int *d_ldu_to_csr_index = nullptr;
int *d_csr_row_index= nullptr;
int *d_csr_col_index= nullptr;

Expand Down
Loading