Skip to content

Adapt pooled stream use to cuda::stream_ref #8529

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

Draft
bdice wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

Adapt pooled stream use to cuda::stream_ref #8529

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
4 changes: 2 additions & 2 deletions cpp/src/randomforest/randomforest.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -363,7 +363,7 @@ class RandomForest {
int stream_id = omp_get_thread_num();
auto s = handle.get_stream_from_stream_pool(stream_id);

auto& selected_rows = row_sampler.sample(i, stream_id, s);
auto& selected_rows = row_sampler.sample(i, stream_id, s.get());

/* Build individual tree in the forest.
- input is a pointer to orig data that have n_cols features and n_rows rows.
Expand All @@ -375,7 +375,7 @@ class RandomForest {
*/

forest->trees[i] = DT::DecisionTree::fit(handle,
s,
s.get(),
input,
n_cols,
n_rows,
Expand Down
4 changes: 2 additions & 2 deletions cpp/src/svm/linear.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
/*
* SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION.
* SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: Apache-2.0
*/
#include <common/nvtx.hpp>
Expand Down Expand Up @@ -112,7 +112,7 @@ class WorkerHandle {
: handle_ptr{new raft::handle_t{h.get_next_usable_stream(stream_id)}},
stream_id(stream_id),
handle(*handle_ptr),
stream(h.get_next_usable_stream(stream_id))
stream(h.get_next_usable_stream(stream_id).get())
{
}

Expand Down
10 changes: 5 additions & 5 deletions cpp/src_prims/selection/knn.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
/*
* SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION.
* SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: Apache-2.0
*/

Expand Down Expand Up @@ -192,7 +192,7 @@ void class_probs(const raft::handle_t& handle,
const float* weights = nullptr)
{
for (std::size_t i = 0; i < y.size(); i++) {
cudaStream_t stream = handle.get_next_usable_stream();
cudaStream_t stream = handle.get_next_usable_stream().get();

int n_unique_labels = n_unique[i];
size_t cur_size = n_query_rows * n_unique_labels;
Expand Down Expand Up @@ -274,7 +274,7 @@ void knn_classify(const raft::handle_t& handle,
for (std::size_t i = 0; i < n_unique.size(); i++) {
int size = n_unique[i];

cudaStream_t stream = handle.get_next_usable_stream(i);
cudaStream_t stream = handle.get_next_usable_stream(i).get();

tmp_probs.emplace_back(n_query_rows * size, stream);
probs.push_back(tmp_probs.back().data());
Expand All @@ -293,7 +293,7 @@ void knn_classify(const raft::handle_t& handle,
dim3 blk(TPB_X, 1, 1);

for (std::size_t i = 0; i < y.size(); i++) {
cudaStream_t stream = handle.get_next_usable_stream(i);
cudaStream_t stream = handle.get_next_usable_stream(i).get();

int n_unique_labels = n_unique[i];

Expand Down Expand Up @@ -347,7 +347,7 @@ void knn_regress(const raft::handle_t& handle,
* Vote average regression value
*/
for (std::size_t i = 0; i < y.size(); i++) {
cudaStream_t stream = handle.get_next_usable_stream();
cudaStream_t stream = handle.get_next_usable_stream().get();

regress_avg_kernel<ValType, precomp_lbls>
<<<raft::ceildiv(n_query_rows, static_cast<std::size_t>(TPB_X)), TPB_X, 0, stream>>>(
Expand Down
45 changes: 24 additions & 21 deletions cpp/tests/sg/rf_test.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -254,13 +254,14 @@ std::shared_ptr<thrust::device_vector<LabelT>> nvForestPredict(
TreeliteModelHandle model;
build_treelite_forest(&model, forest, params.n_cols);

auto nvforest_model = nvforest::import_from_treelite_handle(model,
nvforest::tree_layout::breadth_first,
128,
std::is_same_v<DataT, double>,
nvforest::device_type::gpu,
handle.get_device(),
handle.get_next_usable_stream());
auto nvforest_model =
nvforest::import_from_treelite_handle(model,
nvforest::tree_layout::breadth_first,
128,
std::is_same_v<DataT, double>,
nvforest::device_type::gpu,
handle.get_device(),
handle.get_next_usable_stream().get());
handle.sync_stream();
handle.sync_stream_pool();
delete static_cast<treelite::Model*>(model);
Expand Down Expand Up @@ -325,13 +326,14 @@ auto nvForestPredictProba(const raft::handle_t& handle,
TreeliteModelHandle model;
build_treelite_forest(&model, forest, params.n_cols);

auto nvforest_model = nvforest::import_from_treelite_handle(model,
nvforest::tree_layout::breadth_first,
128,
std::is_same_v<DataT, double>,
nvforest::device_type::gpu,
handle.get_device(),
handle.get_next_usable_stream());
auto nvforest_model =
nvforest::import_from_treelite_handle(model,
nvforest::tree_layout::breadth_first,
128,
std::is_same_v<DataT, double>,
nvforest::device_type::gpu,
handle.get_device(),
handle.get_next_usable_stream().get());
handle.sync_stream();
handle.sync_stream_pool();
delete static_cast<treelite::Model*>(model);
Expand Down Expand Up @@ -929,13 +931,14 @@ TEST(RfTests, IntegerOverflow)
TreeliteModelHandle model;
build_treelite_forest(&model, forest_ptr, n);

auto nvforest_model = nvforest::import_from_treelite_handle(model,
nvforest::tree_layout::breadth_first,
128,
false,
nvforest::device_type::gpu,
handle.get_device(),
handle.get_next_usable_stream());
auto nvforest_model =
nvforest::import_from_treelite_handle(model,
nvforest::tree_layout::breadth_first,
128,
false,
nvforest::device_type::gpu,
handle.get_device(),
handle.get_next_usable_stream().get());
handle.sync_stream();
handle.sync_stream_pool();
delete static_cast<treelite::Model*>(model);
Expand Down
Loading