Add ScopedStream testing for UR CUDA adapter

Author: Hugh Delaney

test/adapters/cuda/CMakeLists.txt

@@ -13,8 +13,12 @@ add_adapter_test(cuda
         urDeviceCreateWithNativeHandle.cpp
         urEventGetNativeHandle.cpp
         urEventCreateWithNativeHandle.cpp
+        urQueueGetNativeHandle.cpp
         kernel_tests.cpp
         memory_tests.cpp
+        #FIXME: make this cleaner
+        ${CMAKE_CURRENT_SOURCE_DIR}/../../../source/adapters/cuda/queue.cpp
+        ${CMAKE_CURRENT_SOURCE_DIR}/../../../source/adapters/cuda/common.cpp
     ENVIRONMENT
         "UR_ADAPTERS_FORCE_LOAD=\"$<TARGET_FILE:ur_adapter_cuda>\""
 )

test/adapters/cuda/urQueueGetNativeHandle.cpp

@@ -0,0 +1,67 @@
+// Copyright (C) 2022-2024 Intel Corporation
+// Part of the Unified-Runtime Project, under the Apache License v2.0 with LLVM Exceptions.
+// See LICENSE.TXT
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+#include "fixtures.h"
+#include "queue.hpp"
+
+using urCudaQueueGetNativeHandleTest = uur::urQueueTest;
+UUR_INSTANTIATE_DEVICE_TEST_SUITE_P(urCudaQueueGetNativeHandleTest);
+
+TEST_P(urCudaQueueGetNativeHandleTest, Success) {
+    CUstream Stream;
+    ASSERT_SUCCESS(
+        urQueueGetNativeHandle(queue, nullptr, (ur_native_handle_t *)&Stream));
+    ASSERT_SUCCESS_CUDA(cuStreamSynchronize(Stream));
+}
+
+TEST_P(urCudaQueueGetNativeHandleTest, OutOfOrder) {
+    CUstream Stream;
+    ur_queue_properties_t props = {
+        /*.stype =*/UR_STRUCTURE_TYPE_QUEUE_PROPERTIES,
+        /*.pNext =*/nullptr,
+        /*.flags =*/UR_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE,
+    };
+    ASSERT_SUCCESS(urQueueCreate(context, device, &props, &queue));
+    ASSERT_SUCCESS(
+        urQueueGetNativeHandle(queue, nullptr, (ur_native_handle_t *)&Stream));
+    ASSERT_SUCCESS_CUDA(cuStreamSynchronize(Stream));
+}
+
+TEST_P(urCudaQueueGetNativeHandleTest, ScopedStream) {
+    CUstream Stream1, Stream2;
+    ur_queue_properties_t props = {
+        /*.stype =*/UR_STRUCTURE_TYPE_QUEUE_PROPERTIES,
+        /*.pNext =*/nullptr,
+        /*.flags =*/UR_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE,
+    };
+    ur_queue_handle_t OutOfOrderQueue;
+    ASSERT_SUCCESS(urQueueCreate(context, device, &props, &OutOfOrderQueue));
+    ASSERT_SUCCESS(urQueueGetNativeHandle(OutOfOrderQueue, nullptr,
+                                          (ur_native_handle_t *)&Stream1));
+    ASSERT_SUCCESS(urQueueGetNativeHandle(OutOfOrderQueue, nullptr,
+                                          (ur_native_handle_t *)&Stream2));
+
+    // We might want to remove this assertion at some point. This is just
+    // testing current implementated behaviour that getting the native OutOfOrderQueue
+    // will call `getNextComputeStream`
+    ASSERT_NE(Stream1, Stream2);
+
+    {
+        ScopedStream ActiveStream(OutOfOrderQueue, 0, nullptr);
+
+        ASSERT_SUCCESS(urQueueGetNativeHandle(OutOfOrderQueue, nullptr,
+                                              (ur_native_handle_t *)&Stream1));
+        ASSERT_SUCCESS(urQueueGetNativeHandle(OutOfOrderQueue, nullptr,
+                                              (ur_native_handle_t *)&Stream2));
+        ASSERT_EQ(Stream1, Stream2);
+    }
+
+    // Go back to returning new streams each time
+    ASSERT_SUCCESS(urQueueGetNativeHandle(OutOfOrderQueue, nullptr,
+                                          (ur_native_handle_t *)&Stream1));
+    ASSERT_SUCCESS(urQueueGetNativeHandle(OutOfOrderQueue, nullptr,
+                                          (ur_native_handle_t *)&Stream2));
+    ASSERT_NE(Stream1, Stream2);
+}