Skip to content
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

Jump to bottom

Use less device memory when checkpointing particles #3238

Merged
merged 9 commits into from
Mar 29, 2024
Merged

Use less device memory when checkpointing particles #3238

Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
4b68202
Use less device memory when checkpointing particles
atmyers Apr 4, 2023
335fcbf
fix warning
atmyers Apr 4, 2023
eb60bf2
Merge branch 'development' into less_device_mem_particle_checkpoint
atmyers Apr 5, 2023
a957a89
Merge branch 'less_device_mem_particle_checkpoint' of github.com:atmy…
atmyers Apr 5, 2023
8c025db
Remove Unnecessary Sync
ax3l Apr 18, 2023
cf5a96e
Merge branch 'development' into less_device_mem_particle_checkpoint
asalmgren Aug 23, 2023
ec50dc6
Update Src/Particle/AMReX_WriteBinaryParticleData.H
atmyers Aug 23, 2023
1bddaf0
Merge branch 'development' into less_device_mem_particle_checkpoint
atmyers Mar 29, 2024
e780848
Merge branch 'less_device_mem_particle_checkpoint' of github.com:atmy…
atmyers Mar 29, 2024
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
21 changes: 11 additions & 10 deletions Src/Particle/AMReX_WriteBinaryParticleData.H
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -190,16 +190,12 @@ packIOData (Vector<int>& idata, Vector<ParticleReal>& rdata, const PC& pc, int l
const Long rChunkSize = AMREX_SPACEDIM + num_output_real;
rdata.resize(np*rChunkSize);

typename PC::IntVector idata_d(idata.size());
typename PC::RealVector rdata_d(rdata.size());

typename PC::IntVector write_int_comp_d(write_int_comp.size());
typename PC::IntVector write_real_comp_d(write_real_comp.size());
Gpu::copyAsync(Gpu::hostToDevice, write_int_comp.begin(), write_int_comp.end(),
write_int_comp_d.begin());
Gpu::copyAsync(Gpu::hostToDevice, write_real_comp.begin(), write_real_comp.end(),
write_real_comp_d.begin());
Gpu::Device::streamSynchronize();

const auto write_int_comp_d_ptr = write_int_comp_d.data();
const auto write_real_comp_d_ptr = write_real_comp_d.data();
Expand All @@ -212,6 +208,9 @@ packIOData (Vector<int>& idata, Vector<ParticleReal>& rdata, const PC& pc, int l
typename PC::IntVector offsets(np_tile);
int num_copies = Scan::ExclusiveSum(np_tile, pflags.begin(), offsets.begin(), Scan::retSum);

typename PC::IntVector idata_d(num_copies*iChunkSize);
typename PC::RealVector rdata_d(num_copies*rChunkSize);

const auto flag_ptr = pflags.data();

auto idata_d_ptr = idata_d.data();
Expand All @@ -225,11 +224,11 @@ packIOData (Vector<int>& idata, Vector<ParticleReal>& rdata, const PC& pc, int l
const auto p = ptd.getSuperParticle(pindex);

if (flag_ptr[pindex]) {
std::size_t iout_index = (pindex+poffset)*iChunkSize;
std::size_t iout_index = pindex*iChunkSize;
packParticleIDs(&idata_d_ptr[iout_index], p, is_checkpoint);
iout_index += 2;

std::size_t rout_index = (pindex+poffset)*rChunkSize;
std::size_t rout_index = pindex*rChunkSize;
for (int j = 0; j < AMREX_SPACEDIM; j++) {
rdata_d_ptr[rout_index] = p.pos(j);
rout_index++;
Expand Down Expand Up @@ -265,12 +264,14 @@ packIOData (Vector<int>& idata, Vector<ParticleReal>& rdata, const PC& pc, int l
}
});

Gpu::copyAsync(Gpu::deviceToHost, idata_d.begin(), idata_d.end(),
idata.begin() + typename PC::IntVector::difference_type(poffset));
Gpu::copyAsync(Gpu::deviceToHost, rdata_d.begin(), rdata_d.end(),
rdata.begin() + typename PC::RealVector::difference_type(poffset));
Gpu::Device::streamSynchronize();

poffset += num_copies;
}

Gpu::copyAsync(Gpu::deviceToHost, idata_d.begin(), idata_d.end(), idata.begin());
Gpu::copyAsync(Gpu::deviceToHost, rdata_d.begin(), rdata_d.end(), rdata.begin());
Gpu::Device::streamSynchronize();
}

template <class PC>
Expand Down
Loading