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

Fix communication for large message sizes #131

Closed
wants to merge 2 commits into from
Closed

Fix communication for large message sizes #131

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6b03a84
Fix communication for large message sizes
masterleinad Oct 7, 2019
c8d2048
Optimize self-communication
masterleinad Oct 8, 2019
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
50 changes: 42 additions & 8 deletions src/details/ArborX_DetailsDistributor.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -179,19 +179,53 @@ class Distributor
int const outdegrees = _destinations.size();
std::vector<MPI_Request> requests;
requests.reserve(outdegrees + indegrees);
int const chunk_size = std::numeric_limits<int>::max();
for (int i = 0; i < indegrees; ++i)
{
requests.emplace_back();
MPI_Irecv(src_buffer.data() + _src_offsets[i] * num_packets,
_src_counts[i] * num_packets * sizeof(ValueType), MPI_BYTE,
_sources[i], MPI_ANY_TAG, _comm, &requests.back());
std::size_t const total_message_size =
_src_counts[i] * num_packets * sizeof(ValueType);
int const n_chunks = (total_message_size + chunk_size - 1) / chunk_size;
if (_sources[i] != comm_rank)
for (int chunk = 0; chunk < n_chunks; ++chunk)
{
requests.emplace_back();
int const this_chunk_size = std::min<std::size_t>(
chunk_size, total_message_size - chunk * chunk_size);
MPI_Irecv(reinterpret_cast<char *>(src_buffer.data() +
_src_offsets[i] * num_packets) +
chunk * chunk_size,
this_chunk_size, MPI_BYTE, _sources[i], 123 + chunk, _comm,
&requests.back());
}
}
for (int i = 0; i < outdegrees; ++i)
{
requests.emplace_back();
MPI_Isend(dest_buffer.data() + _dest_offsets[i] * num_packets,
_dest_counts[i] * num_packets * sizeof(ValueType), MPI_BYTE,
_destinations[i], 123, _comm, &requests.back());
std::size_t const total_message_size =
_dest_counts[i] * num_packets * sizeof(ValueType);
int const n_chunks = (total_message_size + chunk_size - 1) / chunk_size;
if (_destinations[i] == comm_rank)
{
auto const it = std::find(_sources.begin(), _sources.end(), comm_rank);
ARBORX_ASSERT(it != _sources.end());
auto const position = it - _sources.begin();
std::memcpy(src_buffer.data() + _src_offsets[position] * num_packets,
dest_buffer.data() + _dest_offsets[i] * num_packets,
total_message_size);
}
else
{
for (int chunk = 0; chunk < n_chunks; ++chunk)
{
requests.emplace_back();
int const this_chunk_size = std::min<std::size_t>(
chunk_size, total_message_size - chunk * chunk_size);
MPI_Isend(reinterpret_cast<char *>(dest_buffer.data() +
_dest_offsets[i] * num_packets) +
chunk * chunk_size,
this_chunk_size, MPI_BYTE, _destinations[i], 123 + chunk,
_comm, &requests.back());
}
}
}
if (!requests.empty())
MPI_Waitall(requests.size(), requests.data(), MPI_STATUSES_IGNORE);
Expand Down