Skip to content

Remove hostdevice_vector::element due to unnecessary synchronization #19092

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
merged 5 commits into from
Jul 23, 2025
Merged

Remove hostdevice_vector::element due to unnecessary synchronization #19092

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
a18c111
[No Miss-sync] Remove hostdevice_vector::element and replace its only…
JigaoLuo Jun 4, 2025
2103784
[No Miss-sync] Copying inside a refactored lambda
JigaoLuo Jun 6, 2025
8dc770e
Merge branch 'branch-25.08' into no-miss-sync-hostdevicevector_element
mhaseeb123 Jul 22, 2025
b58d179
Addressing Code Review Feedback
JigaoLuo Jul 23, 2025
729a8e9
Merge branch 'branch-25.08' into no-miss-sync-hostdevicevector_element
JigaoLuo Jul 23, 2025
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
12 changes: 11 additions & 1 deletion cpp/src/io/orc/writer_impl.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1146,7 +1146,17 @@ cudf::detail::hostdevice_vector<uint8_t> allocate_and_encode_blobs(
// figure out the buffer size needed for protobuf format
orc_init_statistics_buffersize(
stats_merge_groups.device_ptr(), stat_chunks.data(), num_stat_blobs, stream);
auto max_blobs = stats_merge_groups.element(num_stat_blobs - 1, stream);

// get stats_merge_groups[num_stat_blobs - 1] via a host pinned bounce buffer
auto host_pinned_memory =
cudf::detail::make_pinned_vector_async<statistics_merge_group>(1, stream);
CUDF_CUDA_TRY(cudaMemcpyAsync(host_pinned_memory.data(),
stats_merge_groups.device_ptr(num_stat_blobs - 1),
sizeof(statistics_merge_group),
cudaMemcpyDeviceToHost,
stream.value()));
stream.synchronize();
auto max_blobs = host_pinned_memory.front();

cudf::detail::hostdevice_vector<uint8_t> blobs(max_blobs.start_chunk + max_blobs.num_chunks,
stream);
Expand Down
17 changes: 0 additions & 17 deletions cpp/src/io/utilities/hostdevice_vector.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -100,23 +100,6 @@ class hostdevice_vector {
[[nodiscard]] T* d_end() { return device_ptr(size()); }
[[nodiscard]] T const* d_end() const { return device_ptr(size()); }

/**
* @brief Returns the specified element from device memory
*
* @note This function incurs a device to host memcpy and should be used sparingly.
* @note This function synchronizes `stream`.
*
* @throws rmm::out_of_range exception if `element_index >= size()`
*
* @param element_index Index of the desired element
* @param stream The stream on which to perform the copy
* @return The value of the specified element
*/
[[nodiscard]] T element(std::size_t element_index, rmm::cuda_stream_view stream) const
{
return d_data.element(element_index, stream);
}

operator cudf::host_span<T>() { return host_span<T>{h_data}.subspan(0, size()); }
operator cudf::host_span<T const>() const
{
Expand Down