Skip to content

Commit

Permalink
[pre-commit.ci] auto fixes from pre-commit.com hooks
Browse files Browse the repository at this point in the history 
for more information, see https://pre-commit.ci
  • Loading branch information
@pre-commit-ci
pre-commit-ci[bot] committed Feb 25, 2025
1 parent 3f85174 commit 6aa6e38
Showing 1 changed file with 74 additions and 74 deletions.
148 changes: 74 additions & 74 deletions src/particles/ImpactXParticleContainer.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -142,93 +142,93 @@ namespace impactx
}
}

int nthreads = 1;
int nthreads = 1;
#if defined(AMREX_USE_OMP)
nthreads = omp_get_max_threads();
nthreads = omp_get_max_threads();
#endif

const auto& ba = ParticleBoxArray(lid);
auto n_logical = numTilesInBox(ba[gid], true, tile_size);
const auto& ba = ParticleBoxArray(lid);
auto n_logical = numTilesInBox(ba[gid], true, tile_size);

if (n_logical < nthreads ) {
amrex::Print() << "Too few tiles for the number of OpenMP threads. Parallelization will be poor. \n";
}
if (n_logical < nthreads ) {
amrex::Print() << "Too few tiles for the number of OpenMP threads. Parallelization will be poor. \n";
}

for (int ithr = 0; ithr < nthreads; ++ithr) {
DefineAndReturnParticleTile(lid, gid, ithr);
}
for (int ithr = 0; ithr < nthreads; ++ithr) {
DefineAndReturnParticleTile(lid, gid, ithr);
}

int pid = ParticleType::NextID();
ParticleType::NextID(pid+np);
int pid = ParticleType::NextID();
ParticleType::NextID(pid+np);
AMREX_ALWAYS_ASSERT_WITH_MESSAGE(
static_cast<amrex::Long>(pid) + static_cast<amrex::Long>(np) < amrex::LongParticleIds::LastParticleID,
"ERROR: overflow on particle id numbers");

#if defined(AMREX_USE_OMP)
#pragma omp parallel if (amrex::Gpu::notInLaunchRegion())
#endif
{
int tid = omp_get_thread_num();
int nr = np / nthreads;
int nlft = np - nr*nthreads;

int num_to_add = 0;
int my_index =0;
if (tid < nlft) { // get nr+1 items
my_index = tid * (nr+1);
num_to_add = nr+1;
} else { // get nr items
my_index = tid * nr + nlft;
num_to_add = nr;
}

auto& particle_tile = ParticlesAt(lid, gid, tid);
auto old_np = particle_tile.numParticles();
auto new_np = old_np + num_to_add;
particle_tile.resize(new_np);

const int cpuid = amrex::ParallelDescriptor::MyProc();

auto & soa = particle_tile.GetStructOfArrays().GetRealData();
amrex::ParticleReal * const AMREX_RESTRICT x_arr = soa[RealSoA::x].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT y_arr = soa[RealSoA::y].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT t_arr = soa[RealSoA::t].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT px_arr = soa[RealSoA::px].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT py_arr = soa[RealSoA::py].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT pt_arr = soa[RealSoA::pt].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT qm_arr = soa[RealSoA::qm].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT w_arr = soa[RealSoA::w ].dataPtr();

uint64_t * const AMREX_RESTRICT idcpu_arr = particle_tile.GetStructOfArrays().GetIdCPUData().dataPtr();

amrex::ParticleReal const * const AMREX_RESTRICT x_ptr = x.data();
amrex::ParticleReal const * const AMREX_RESTRICT y_ptr = y.data();
amrex::ParticleReal const * const AMREX_RESTRICT t_ptr = t.data();
amrex::ParticleReal const * const AMREX_RESTRICT px_ptr = px.data();
amrex::ParticleReal const * const AMREX_RESTRICT py_ptr = py.data();
amrex::ParticleReal const * const AMREX_RESTRICT pt_ptr = pt.data();

amrex::ParallelFor(num_to_add,
[=] AMREX_GPU_DEVICE (int i) noexcept
{
idcpu_arr[old_np+i] = amrex::SetParticleIDandCPU(pid + my_index + i, cpuid);

x_arr[old_np+i] = x_ptr[my_index+i];
y_arr[old_np+i] = y_ptr[my_index+i];
t_arr[old_np+i] = t_ptr[my_index+i];

px_arr[old_np+i] = px_ptr[my_index+i];
py_arr[old_np+i] = py_ptr[my_index+i];
pt_arr[old_np+i] = pt_ptr[my_index+i];

qm_arr[old_np+i] = qm;
w_arr[old_np+i] = bchchg/ablastr::constant::SI::q_e/np;
});
}

// safety first: in case passed attribute arrays were temporary, we
// want to make sure the ParallelFor has ended here
amrex::Gpu::streamSynchronize();
{
int tid = omp_get_thread_num();
int nr = np / nthreads;
int nlft = np - nr*nthreads;

int num_to_add = 0;
int my_index =0;
if (tid < nlft) { // get nr+1 items
my_index = tid * (nr+1);
num_to_add = nr+1;
} else { // get nr items
my_index = tid * nr + nlft;
num_to_add = nr;
}

auto& particle_tile = ParticlesAt(lid, gid, tid);
auto old_np = particle_tile.numParticles();
auto new_np = old_np + num_to_add;
particle_tile.resize(new_np);

const int cpuid = amrex::ParallelDescriptor::MyProc();

auto & soa = particle_tile.GetStructOfArrays().GetRealData();
amrex::ParticleReal * const AMREX_RESTRICT x_arr = soa[RealSoA::x].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT y_arr = soa[RealSoA::y].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT t_arr = soa[RealSoA::t].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT px_arr = soa[RealSoA::px].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT py_arr = soa[RealSoA::py].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT pt_arr = soa[RealSoA::pt].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT qm_arr = soa[RealSoA::qm].dataPtr();
amrex::ParticleReal * const AMREX_RESTRICT w_arr = soa[RealSoA::w ].dataPtr();

uint64_t * const AMREX_RESTRICT idcpu_arr = particle_tile.GetStructOfArrays().GetIdCPUData().dataPtr();

amrex::ParticleReal const * const AMREX_RESTRICT x_ptr = x.data();
amrex::ParticleReal const * const AMREX_RESTRICT y_ptr = y.data();
amrex::ParticleReal const * const AMREX_RESTRICT t_ptr = t.data();
amrex::ParticleReal const * const AMREX_RESTRICT px_ptr = px.data();
amrex::ParticleReal const * const AMREX_RESTRICT py_ptr = py.data();
amrex::ParticleReal const * const AMREX_RESTRICT pt_ptr = pt.data();

amrex::ParallelFor(num_to_add,
[=] AMREX_GPU_DEVICE (int i) noexcept
{
idcpu_arr[old_np+i] = amrex::SetParticleIDandCPU(pid + my_index + i, cpuid);

x_arr[old_np+i] = x_ptr[my_index+i];
y_arr[old_np+i] = y_ptr[my_index+i];
t_arr[old_np+i] = t_ptr[my_index+i];

px_arr[old_np+i] = px_ptr[my_index+i];
py_arr[old_np+i] = py_ptr[my_index+i];
pt_arr[old_np+i] = pt_ptr[my_index+i];

qm_arr[old_np+i] = qm;
w_arr[old_np+i] = bchchg/ablastr::constant::SI::q_e/np;
});
}

// safety first: in case passed attribute arrays were temporary, we
// want to make sure the ParallelFor has ended here
amrex::Gpu::streamSynchronize();
}

void
Expand Down

0 comments on commit 6aa6e38

Please sign in to comment.