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

Optimize ORC reader performance for list data #13708

Merged
merged 6 commits into from
Jul 25, 2023
Merged

Optimize ORC reader performance for list data #13708

Changes from 3 commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
e99e365
One scan per column (with lots of commented code from testing)
vyasr Jul 15, 2023
561db9d
cleanup
vyasr Jul 15, 2023
f2c4a38
Remove unnecessary sync
vyasr Jul 15, 2023
29f987f
Merge remote-tracking branch 'origin/branch-23.08' into fix/issue_13674
vyasr Jul 19, 2023
c8de39c
Switch to host span to remove round-trip
vyasr Jul 19, 2023
48e5465
PR reviews
vyasr Jul 24, 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
14 changes: 9 additions & 5 deletions cpp/src/io/orc/reader_impl.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -725,12 +725,16 @@ struct list_buffer_data {
void generate_offsets_for_list(device_span<list_buffer_data const> buff_data,
vyasr marked this conversation as resolved.
Show resolved Hide resolved
rmm::cuda_stream_view stream)
{
auto const transformer = [] __device__(list_buffer_data const list_data) {
std::vector<list_buffer_data> buff_data_vec(buff_data.size());
CUDF_CUDA_TRY(cudaMemcpyAsync(buff_data_vec.data(),
buff_data.data(),
sizeof(list_buffer_data) * buff_data.size(),
cudaMemcpyDeviceToHost,
vyasr marked this conversation as resolved.
Show resolved Hide resolved
stream.value()));
for (auto list_data : buff_data_vec) {
thrust::exclusive_scan(
thrust::seq, list_data.data, list_data.data + list_data.size, list_data.data);
};
thrust::for_each(rmm::exec_policy(stream), buff_data.begin(), buff_data.end(), transformer);
stream.synchronize();
rmm::exec_policy(stream), list_data.data, list_data.data + list_data.size, list_data.data);
vyasr marked this conversation as resolved.
Show resolved Hide resolved
}
}

/**
Expand Down