From 7dde2f67108b3626c2fffb8422950add4cb5ebee Mon Sep 17 00:00:00 2001
From: Simone <simone.balducci00@gmail.com>
Date: Thu, 18 Dec 2025 19:28:53 +0100
Subject: [PATCH 1/2] Implement `get_queue` overload for native streams

---
 include/CLUEstering/utils/get_queue.hpp | 24 ++++++++++++---
 tests/test_utilities.cpp                | 40 +++++++++++++++++++++++++
 2 files changed, 60 insertions(+), 4 deletions(-)

diff --git a/include/CLUEstering/utils/get_queue.hpp b/include/CLUEstering/utils/get_queue.hpp
index 02300332..5b71220d 100644
--- a/include/CLUEstering/utils/get_queue.hpp
+++ b/include/CLUEstering/utils/get_queue.hpp
@@ -17,9 +17,9 @@ namespace clue {
   /// @param device_id The index of the device
   /// @return An alpaka queue created from the device corresponding to the given index
   template <std::integral TIdx>
-  inline clue::Queue get_queue(TIdx device_id = TIdx{}) {
+  inline auto get_queue(TIdx device_id = TIdx{}) {
     auto device = alpaka::getDevByIdx(clue::Platform{}, device_id);
-    return clue::Queue{device};
+    return clue::Queue(device);
   }
 
   /// @brief Get an alpaka queue created from a given device
@@ -28,8 +28,24 @@ namespace clue {
   /// @param device The device to create the queue from
   /// @return An alpaka queue created from the given device
   template <concepts::device TDevice>
-  inline clue::Queue get_queue(const TDevice& device) {
-    return clue::Queue{device};
+  inline auto get_queue(const TDevice& device) {
+    return clue::Queue(device);
   }
 
+#ifdef ALPAKA_ACC_GPU_CUDA_ENABLED
+  /// @brief Get an alpaka queue wrapping a CUDA stream
+  ///
+  /// @param stream The CUDA stream to wrap inside the alpaka queue
+  /// @return An alpaka queue wrapping the given CUDA stream
+  inline auto get_queue(cudaStream_t& stream) { return clue::Queue(stream); }
+#endif
+
+#ifdef ALPAKA_ACC_GPU_HIP_ENABLED
+  /// @brief Get an alpaka queue wrapping a HIP stream
+  ///
+  /// @param stream The HIP stream to wrap inside the alpaka queue
+  /// @return An alpaka queue wrapping the given HIP stream
+  inline auto get_queue(hipStream_t& stream) { return clue::Queue(stream); }
+#endif
+
 }  // namespace clue
diff --git a/tests/test_utilities.cpp b/tests/test_utilities.cpp
index 328b5e6e..b4519edf 100644
--- a/tests/test_utilities.cpp
+++ b/tests/test_utilities.cpp
@@ -58,6 +58,46 @@ TEST_CASE("Test clue::get_queue utility") {
     auto d_points2 = clue::PointsDevice<2>(queue2, points2.size());
     CHECK(1);
   }
+
+#ifdef ALPAKA_ACC_GPU_CUDA_ENABLED
+  SUBCASE("Create queue using CUDA stream") {
+    cudaStream_t stream;
+    cudaStreamCreate(&stream);
+
+    {
+      auto queue = clue::get_queue(stream);
+      static_assert(std::is_same_v<decltype(queue), clue::Queue>, "Expected type clue::Queue");
+      CHECK(alpaka::getDev(queue) == alpaka::getDevByIdx(clue::Platform{}, 0u));
+
+      // check if data allocation works
+      clue::PointsHost<2> points1(queue, 1000);
+      auto d_points1 = clue::PointsDevice<2>(queue, points1.size());
+      CHECK(1);
+    }
+
+    cudaStreamDestroy(stream);
+  }
+#endif
+
+#ifdef ALPAKA_ACC_GPU_HIP_ENABLED
+  SUBCASE("Create queue using HIP stream") {
+    hipStream_t stream;
+    hipStreamCreate(&stream);
+
+    {
+      auto queue = clue::get_queue(stream);
+      static_assert(std::is_same_v<decltype(queue), clue::Queue>, "Expected type clue::Queue");
+      CHECK(alpaka::getDev(queue) == alpaka::getDevByIdx(clue::Platform{}, 0u));
+
+      // check if data allocation works
+      clue::PointsHost<2> points1(queue, 1000);
+      auto d_points1 = clue::PointsDevice<2>(queue, points1.size());
+      CHECK(1);
+    }
+
+    hipStreamDestroy(stream);
+  }
+#endif
 }
 
 TEST_CASE("Test get_clusters host function") {

From bd5e70fa13db1b4d638b73a0000978cb767f372a Mon Sep 17 00:00:00 2001
From: Simone <simone.balducci00@gmail.com>
Date: Thu, 18 Dec 2025 19:29:59 +0100
Subject: [PATCH 2/2] Update native examples

---
 examples/cuda_native/main.cu | 2 +-
 examples/hip_native/main.cpp | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/examples/cuda_native/main.cu b/examples/cuda_native/main.cu
index e7d04d24..bec2cb07 100644
--- a/examples/cuda_native/main.cu
+++ b/examples/cuda_native/main.cu
@@ -23,7 +23,7 @@ void compute_clusters(TQueue& queue, std::vector<int>& cluster_indexes) {
 int main() {
   cudaStream_t stream;
   cudaStreamCreate(&stream);
-  clue::Queue queue(stream);
+  auto queue = clue::get_queue(stream);
 
   std::vector<int> cluster_indexes;
   compute_clusters(queue, cluster_indexes);
diff --git a/examples/hip_native/main.cpp b/examples/hip_native/main.cpp
index 65aec4e0..0a214e98 100644
--- a/examples/hip_native/main.cpp
+++ b/examples/hip_native/main.cpp
@@ -23,7 +23,7 @@ void compute_clusters(TQueue& queue, std::vector<int>& cluster_indexes) {
 int main() {
   hipStream_t stream;
   hipStreamCreate(&stream);
-  clue::Queue queue(stream);
+  auto queue = clue::get_queue(stream);
 
   std::vector<int> cluster_indexes;
   compute_clusters(queue, cluster_indexes);