add IPC support to default async mempool

Author: ksimpson-work

cuda_bindings/cuda/bindings/driver.pyx.in

@@ -34305,7 +34305,7 @@ def cuMemPrefetchAsync_v2(devPtr, size_t count, location not None : CUmemLocatio
     to the host NUMA node closest to the current thread's CPU by specifying
     :py:obj:`~.CU_MEM_LOCATION_TYPE_HOST_NUMA_CURRENT` for
     :py:obj:`~.CUmemLocation.type`. Note when
-    :py:obj:`~.CUmemLocation.type` is etiher
+    :py:obj:`~.type` is etiher
     :py:obj:`~.CU_MEM_LOCATION_TYPE_HOST` OR
     :py:obj:`~.CU_MEM_LOCATION_TYPE_HOST_NUMA_CURRENT`,
     :py:obj:`~.CUmemLocation.id` will be ignored.

cuda_core/cuda/core/experimental/__init__.py

@@ -7,6 +7,7 @@
 from cuda.core.experimental._event import EventOptions
 from cuda.core.experimental._launcher import LaunchConfig, launch
 from cuda.core.experimental._linker import Linker, LinkerOptions
+from cuda.core.experimental._memory import AsyncMempool
 from cuda.core.experimental._module import ObjectCode
 from cuda.core.experimental._program import Program, ProgramOptions
 from cuda.core.experimental._stream import Stream, StreamOptions

cuda_core/cuda/core/experimental/_device.py

@@ -6,7 +6,7 @@
 from typing import Union
 
 from cuda.core.experimental._context import Context, ContextOptions
-from cuda.core.experimental._memory import Buffer, MemoryResource, _DefaultAsyncMempool, _SynchronousMemoryResource
+from cuda.core.experimental._memory import AsyncMempool, Buffer, MemoryResource, _SynchronousMemoryResource
 from cuda.core.experimental._stream import Stream, StreamOptions, default_stream
 from cuda.core.experimental._utils import ComputeCapability, CUDAError, driver, handle_return, precondition, runtime
 
@@ -962,7 +962,7 @@ def __new__(cls, device_id=None):
                             runtime.cudaDeviceGetAttribute(runtime.cudaDeviceAttr.cudaDevAttrMemoryPoolsSupported, 0)
                         )
                     ) == 1:
-                        dev._mr = _DefaultAsyncMempool(dev_id)
+                        dev._mr = AsyncMempool._from_device(dev_id)
                     else:
                         dev._mr = _SynchronousMemoryResource(dev_id)