Address code review feedback: move skip logic, add docstring note, update release note

Co-authored-by: kkraus14 <3665167+kkraus14@users.noreply.github.com>

Author: Copilot

cuda_core/cuda/core/experimental/_memory.py

@@ -318,6 +318,13 @@ class DeviceMemoryResource(MemoryResource):
         Device ordinal for which a memory resource is constructed. The mempool that is
         set to *current* on ``device_id`` is used. If no mempool is set to current yet,
         the driver would use the *default* mempool on the device.
+
+    Notes
+    -----
+    During initialization, this class automatically configures the memory pool's release
+    threshold for improved performance. If the current threshold is 0 (default), it is
+    set to the maximum value to prevent immediate memory release when there are no active
+    allocations, which can cause performance degradation.
     """
 
     __slots__ = ("_dev_id",)

cuda_core/docs/source/release/0.X.Y-notes.rst

@@ -36,5 +36,5 @@ None.
 Fixes and enhancements
 ----------------------
 
-- Improved :class:`DeviceMemoryResource` allocation performance when there are no active allocations (addresses issue #771).
+- Improved :class:`DeviceMemoryResource` allocation performance when there are no active allocations by setting a higher release threshold (addresses issue #771).
 - Fix :class:`LaunchConfig` grid unit conversion when cluster is set (addresses issue #867).

cuda_core/tests/test_memory.py

@@ -259,14 +259,15 @@ def test_buffer_dunder_dlpack_device_failure():
         buffer.__dlpack_device__()
 
 
-@pytest.mark.skipif(not Device().properties.memory_pools_supported, reason="memory pools not supported")
 def test_device_memory_resource_initialization():
     """Test that DeviceMemoryResource can be initialized successfully.
 
     This test verifies that the DeviceMemoryResource initializes properly,
     including the release threshold configuration for performance optimization.
     """
     device = Device()
+    if not device.properties.memory_pools_supported:
+        pytest.skip("memory pools not supported")
     device.set_current()
 
     # This should succeed and configure the memory pool release threshold