Skip to content

Simplifying AAH sweep chunks to reduce duplicated code #898

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
wdhawkins merged 1 commit into from
Jan 30, 2026
Merged

Simplifying AAH sweep chunks to reduce duplicated code #898

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
184 changes: 144 additions & 40 deletions ...s/linear_boltzmann_solvers/discrete_ordinates_problem/sweep_chunks/aah_avx_sweep_chunk.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
// SPDX-License-Identifier: MIT

#include "modules/linear_boltzmann_solvers/discrete_ordinates_problem/sweep_chunks/aah_sweep_chunk.h"
#include "modules/linear_boltzmann_solvers/discrete_ordinates_problem/sweep_chunks/aah_sweep_kernels.h"
#include "modules/linear_boltzmann_solvers/discrete_ordinates_problem/sweep/fluds/aah_fluds.h"
#include "framework/mesh/mesh_continuum/mesh_continuum.h"
#include "framework/logging/log_exceptions.h"
Expand Down Expand Up @@ -207,42 +208,42 @@ inline void static SimdBatchSolve(const double* Am,

} // namespace detail

template <int NumNodes>
template <int NumNodes, bool time_dependent>
void
AAHSweepChunk::CPUSweep_FixedN(AngleSet& angle_set)
AAH_Sweep_FixedN(AAHSweepData& data, AngleSet& angle_set)
{
static_assert(NumNodes >= 1 and NumNodes <= 8);

CALI_CXX_MARK_SCOPE("AAHSweepChunk::CPUSweep_FixedN");
CALI_CXX_MARK_SCOPE("AAH_Sweep_FixedN");

const size_t gs_size = groupset_.groups.size();
const auto gs_gi = groupset_.groups.front().id;
const auto& groupset = data.groupset;
const size_t gs_size = groupset.groups.size();
const auto gs_gi = groupset.groups.front().id;

int deploc_face_counter = -1;
int preloc_face_counter = -1;

auto& fluds = dynamic_cast<AAH_FLUDS&>(angle_set.GetFLUDS());
const auto& m2d_op = groupset_.quadrature->GetMomentToDiscreteOperator();
const auto& d2m_op = groupset_.quadrature->GetDiscreteToMomentOperator();
const auto& m2d_op = groupset.quadrature->GetMomentToDiscreteOperator();
const auto& d2m_op = groupset.quadrature->GetDiscreteToMomentOperator();

std::vector<double> b(gs_size * NumNodes, 0.0);
std::vector<double> sigma_block;
sigma_block.reserve(group_block_size_);
sigma_block.reserve(data.group_block_size);

const auto& spds = angle_set.GetSPDS();
const auto& spls = spds.GetLocalSubgrid();
const size_t num_spls = spls.size();
for (size_t spls_index = 0; spls_index < num_spls; ++spls_index)
{
const uint64_t cell_local_id = spls[spls_index];
auto& cell = grid_->local_cells[cell_local_id];
auto& cell_transport_view = cell_transport_views_[cell_local_id];
const auto& cell_mapping = discretization_.GetCellMapping(cell);
auto& cell = data.grid->local_cells[cell_local_id];
auto& cell_transport_view = data.cell_transport_views[cell_local_id];
const auto& cell_mapping = data.discretization.GetCellMapping(cell);
const size_t cell_num_nodes = cell_mapping.GetNumNodes();
constexpr auto expected_nodes = static_cast<size_t>(NumNodes);
OpenSnInvalidArgumentIf(
cell_num_nodes != expected_nodes,
"AAHSweepChunk::CPUSweep_FixedN invoked for an incompatible cell topology.");
OpenSnInvalidArgumentIf(cell_num_nodes != expected_nodes,
"AAH_Sweep_FixedN invoked for an incompatible cell topology.");

const auto& face_orientations = spds.GetCellFaceOrientations()[cell_local_id];
const size_t cell_num_faces = cell.faces.size();
Expand All @@ -251,10 +252,10 @@ AAHSweepChunk::CPUSweep_FixedN(AngleSet& angle_set)
const int ni_deploc_face_counter = deploc_face_counter;
const int ni_preloc_face_counter = preloc_face_counter;

const double rho = densities_[cell.local_id];
const auto& sigma_t = xs_.at(cell.block_id)->GetSigmaTotal();
const double rho = data.densities[cell.local_id];
const auto& sigma_t = data.xs.at(cell.block_id)->GetSigmaTotal();

const auto& unit_mats = unit_cell_matrices_[cell_local_id];
const auto& unit_mats = data.unit_cell_matrices[cell_local_id];
const auto& G = unit_mats.intV_shapeI_gradshapeJ;
const auto& M = unit_mats.intV_shapeI_shapeJ;
const auto& M_surf = unit_mats.intS_shapeI_shapeJ;
Expand All @@ -273,20 +274,33 @@ AAHSweepChunk::CPUSweep_FixedN(AngleSet& angle_set)
}

std::array<double, matrix_size> Amat{};
std::vector<std::array<size_t, NumNodes>> moment_dof_map(num_moments_);
for (std::size_t m = 0; m < num_moments_; ++m)
std::vector<std::array<size_t, NumNodes>> moment_dof_map(data.num_moments);
for (size_t m = 0; m < data.num_moments; ++m)
{
PRAGMA_UNROLL
for (int i = 0; i < NumNodes; ++i)
moment_dof_map[m][i] = cell_transport_view.MapDOF(i, m, gs_gi);
}

std::vector<double> tau_gsg;
if constexpr (time_dependent)
{
const auto& inv_velg = data.xs.at(cell.block_id)->GetInverseVelocity();
const double theta = data.problem.GetTheta();
const double inv_theta = 1.0 / theta;
const double dt = data.problem.GetTimeStep();
const double inv_dt = 1.0 / dt;
tau_gsg.assign(gs_size, 0.0);
for (size_t gsg = 0; gsg < gs_size; ++gsg)
tau_gsg[gsg] = inv_velg[gs_gi + gsg] * inv_theta * inv_dt;
}

const std::vector<std::uint32_t>& as_angle_indices = angle_set.GetAngleIndices();
for (size_t as_ss_idx = 0; as_ss_idx < as_angle_indices.size(); ++as_ss_idx)
{
const auto direction_num = as_angle_indices[as_ss_idx];
const auto omega = groupset_.quadrature->omegas[direction_num];
const auto wt = groupset_.quadrature->weights[direction_num];
const auto omega = groupset.quadrature->omegas[direction_num];
const auto wt = groupset.quadrature->weights[direction_num];

deploc_face_counter = ni_deploc_face_counter;
preloc_face_counter = ni_preloc_face_counter;
Expand Down Expand Up @@ -339,7 +353,7 @@ AAHSweepChunk::CPUSweep_FixedN(AngleSet& angle_set)
f,
fj,
gs_gi,
IsSurfaceSourceActive());
data.surface_source_active);

for (size_t fi = 0; fi < num_face_nodes; ++fi)
{
Expand All @@ -359,25 +373,32 @@ AAHSweepChunk::CPUSweep_FixedN(AngleSet& angle_set)
const double* __restrict m2d_row = m2d_op[direction_num].data();
const double* __restrict d2m_row = d2m_op[direction_num].data();

for (size_t g0 = 0; g0 < gs_size; g0 += group_block_size_)
const double* psi_old =
(time_dependent and data.psi_old)
? &(*data.psi_old)[data.discretization.MapDOFLocal(cell, 0, groupset.psi_uk_man_, 0, 0)]
: nullptr;

for (size_t g0 = 0; g0 < gs_size; g0 += data.group_block_size)
{
const size_t g1 = std::min(g0 + group_block_size_, gs_size);
const size_t g1 = std::min(g0 + data.group_block_size, gs_size);
const size_t block_len = g1 - g0;
sigma_block.resize(block_len);

for (size_t gsg = g0; gsg < g1; ++gsg)
{
const size_t rel = gsg - g0;
const double sigma_tg = rho * sigma_t[gs_gi + gsg];
double sigma_tg = rho * sigma_t[gs_gi + gsg];
if constexpr (time_dependent)
sigma_tg += tau_gsg[gsg];
sigma_block[rel] = sigma_tg;

double* __restrict bg = &b[gsg * NumNodes];
for (std::size_t m = 0; m < num_moments_; ++m)
for (size_t m = 0; m < data.num_moments; ++m)
{
const double w = m2d_row[m];
std::array<double, NumNodes> nodal_source{};
for (int i = 0; i < NumNodes; ++i)
nodal_source[i] = w * source_moments_[moment_dof_map[m][i] + gsg];
nodal_source[i] = w * data.source_moments[moment_dof_map[m][i] + gsg];

for (int i = 0; i < NumNodes; ++i)
{
Expand All @@ -391,6 +412,33 @@ AAHSweepChunk::CPUSweep_FixedN(AngleSet& angle_set)
}
}

if constexpr (time_dependent)
{
if (data.include_rhs_time_term and psi_old)
{
for (size_t gsg = g0; gsg < g1; ++gsg)
{
const double tau = tau_gsg[gsg];
double* __restrict bg = &b[gsg * NumNodes];

for (int i = 0; i < NumNodes; ++i)
{
double value = 0.0;
const double* row = &mass_matrix[idx(i, 0)];
PRAGMA_UNROLL
for (int j = 0; j < NumNodes; ++j)
{
const size_t imap = static_cast<size_t>(j) * data.groupset_angle_group_stride +
direction_num * data.groupset_group_stride;
const double psi_old_val = psi_old[imap + gsg];
value += row[j] * psi_old_val;
}
bg[i] += tau * value;
}
}
}
}

size_t k = 0;

#if __AVX512F__
Expand Down Expand Up @@ -444,30 +492,42 @@ AAHSweepChunk::CPUSweep_FixedN(AngleSet& angle_set)
for (size_t gsg = g0; gsg < g1; ++gsg)
{
const double* __restrict bg = &b[gsg * NumNodes];
for (std::size_t m = 0; m < num_moments_; ++m)
for (size_t m = 0; m < data.num_moments; ++m)
{
const double w = d2m_row[m];
PRAGMA_UNROLL
for (int i = 0; i < NumNodes; ++i)
{
const size_t dof = cell_transport_view.MapDOF(i, m, gs_gi);
destination_phi_[dof + gsg] += w * bg[i];
data.destination_phi[dof + gsg] += w * bg[i];
}
}
}
}

if (save_angular_flux_)
if (data.save_angular_flux)
{
double* cell_psi_data =
&destination_psi_[discretization_.MapDOFLocal(cell, 0, groupset_.psi_uk_man_, 0, 0)];
&data
.destination_psi[data.discretization.MapDOFLocal(cell, 0, groupset.psi_uk_man_, 0, 0)];
PRAGMA_UNROLL
for (int i = 0; i < NumNodes; ++i)
{
const size_t imap =
i * groupset_angle_group_stride_ + direction_num * groupset_group_stride_;
i * data.groupset_angle_group_stride + direction_num * data.groupset_group_stride;
for (size_t gsg = 0; gsg < gs_size; ++gsg)
cell_psi_data[imap + gsg] = b[gsg * NumNodes + i];
{
const double psi_sol = b[gsg * NumNodes + i];
if constexpr (time_dependent)
{
const double theta = data.problem.GetTheta();
const double inv_theta = 1.0 / theta;
const double psi_old_val = psi_old ? psi_old[imap + gsg] : 0.0;
cell_psi_data[imap + gsg] = inv_theta * (psi_sol + (theta - 1.0) * psi_old_val);
}
else
cell_psi_data[imap + gsg] = psi_sol;
}
}
}

Expand Down Expand Up @@ -523,12 +583,56 @@ AAHSweepChunk::CPUSweep_FixedN(AngleSet& angle_set)
}
}

template void AAHSweepChunk::CPUSweep_FixedN<2>(AngleSet&);
template void AAHSweepChunk::CPUSweep_FixedN<3>(AngleSet&);
template void AAHSweepChunk::CPUSweep_FixedN<4>(AngleSet&);
template void AAHSweepChunk::CPUSweep_FixedN<5>(AngleSet&);
template void AAHSweepChunk::CPUSweep_FixedN<6>(AngleSet&);
template void AAHSweepChunk::CPUSweep_FixedN<7>(AngleSet&);
template void AAHSweepChunk::CPUSweep_FixedN<8>(AngleSet&);
template <int NumNodes>
void
AAHSweepChunk::Sweep_FixedN(AngleSet& angle_set)
{
AAHSweepData data{grid_,
discretization_,
unit_cell_matrices_,
cell_transport_views_,
densities_,
source_moments_,
groupset_,
xs_,
num_moments_,
max_num_cell_dofs_,
min_num_cell_dofs_,
save_angular_flux_,
groupset_angle_group_stride_,
groupset_group_stride_,
destination_phi_,
destination_psi_,
surface_source_active_,
include_rhs_time_term_,
problem_,
nullptr,
group_block_size_};

AAH_Sweep_FixedN<NumNodes, false>(data, angle_set);
}

template void AAH_Sweep_FixedN<2, false>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<3, false>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<4, false>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<5, false>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<6, false>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<7, false>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<8, false>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<2, true>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<3, true>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<4, true>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<5, true>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<6, true>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<7, true>(AAHSweepData&, AngleSet&);
template void AAH_Sweep_FixedN<8, true>(AAHSweepData&, AngleSet&);

template void AAHSweepChunk::Sweep_FixedN<2>(AngleSet&);
template void AAHSweepChunk::Sweep_FixedN<3>(AngleSet&);
template void AAHSweepChunk::Sweep_FixedN<4>(AngleSet&);
template void AAHSweepChunk::Sweep_FixedN<5>(AngleSet&);
template void AAHSweepChunk::Sweep_FixedN<6>(AngleSet&);
template void AAHSweepChunk::Sweep_FixedN<7>(AngleSet&);
template void AAHSweepChunk::Sweep_FixedN<8>(AngleSet&);

} // namespace opensn
Loading