renamed dynamic spilling => object spilling

dask_cuda/cli/dask_cuda_worker.py

@@ -197,9 +197,9 @@
     help="Enable just-in-time unspilling",
 )
 @click.option(
-    "--enable-dynamic-spill/--disable-dynamic-spill",
+    "--enable-object-spilling/--disable-object-spilling",
     default=None,  # If not specified, use Dask config
-    help="Enable dynamic spilling",
+    help="Enable object spilling",
 )
 def main(
     scheduler,
@@ -229,7 +229,7 @@ def main(
     enable_rdmacm,
     net_devices,
     enable_jit_unspill,
-    enable_dynamic_spill,
+    enable_object_spilling,
     **kwargs,
 ):
     if tls_ca_file and tls_cert and tls_key:
@@ -267,7 +267,7 @@ def main(
         enable_rdmacm,
         net_devices,
         enable_jit_unspill,
-        enable_dynamic_spill,
+        enable_object_spilling,
         **kwargs,
     )
 

dask_cuda/cuda_worker.py

@@ -19,7 +19,7 @@
 from distributed.worker import parse_memory_limit
 
 from .device_host_file import DeviceHostFile
-from .dynamic_host_file import DynamicHostFile
+from .object_spilling_host_file import ObjectSpillingHostFile
 from .initialize import initialize
 from .utils import (
     CPUAffinity,
@@ -73,7 +73,7 @@ def __init__(
         enable_rdmacm=False,
         net_devices=None,
         jit_unspill=None,
-        dynamic_spill=None,
+        object_spilling=None,
         **kwargs,
     ):
         # Required by RAPIDS libraries (e.g., cuDF) to ensure no context
@@ -184,12 +184,12 @@ def del_pid_file():
         else:
             self.jit_unspill = jit_unspill
 
-        if dynamic_spill is None:
-            self.dynamic_spill = dask.config.get("dynamic-spill", default=False)
+        if object_spilling is None:
+            self.object_spilling = dask.config.get("object-spilling", default=False)
         else:
-            self.dynamic_spill = dynamic_spill
+            self.object_spilling = object_spilling
 
-        hostfile = DynamicHostFile if self.dynamic_spill else DeviceHostFile
+        hostfile = ObjectSpillingHostFile if self.object_spilling else DeviceHostFile
 
         self.nannies = [
             Nanny(

dask_cuda/local_cuda_cluster.py

@@ -8,7 +8,7 @@
 from distributed.worker import parse_memory_limit
 
 from .device_host_file import DeviceHostFile
-from .dynamic_host_file import DynamicHostFile
+from .object_spilling_host_file import ObjectSpillingHostFile
 from .initialize import initialize
 from .utils import (
     CPUAffinity,
@@ -137,7 +137,7 @@ def __init__(
         rmm_pool_size=None,
         rmm_managed_memory=False,
         jit_unspill=None,
-        dynamic_spill=None,
+        object_spilling=None,
         **kwargs,
     ):
         # Required by RAPIDS libraries (e.g., cuDF) to ensure no context
@@ -192,12 +192,12 @@ def __init__(
         else:
             self.jit_unspill = jit_unspill
 
-        if dynamic_spill is None:
-            self.dynamic_spill = dask.config.get("dynamic-spill", default=False)
+        if object_spilling is None:
+            self.object_spilling = dask.config.get("object-spilling", default=False)
         else:
-            self.dynamic_spill = dynamic_spill
+            self.object_spilling = object_spilling
 
-        hostfile = DynamicHostFile if self.dynamic_spill else DeviceHostFile
+        hostfile = ObjectSpillingHostFile if self.object_spilling else DeviceHostFile
 
         if data is None:
             data = (

dask_cuda/dynamic_host_file.py → dask_cuda/object_spilling_host_file.py

@@ -8,7 +8,7 @@
 from .get_device_memory_objects import get_device_memory_objects
 
 
-class DynamicHostFile(MutableMapping):
+class ObjectSpillingHostFile(MutableMapping):
     """Manages serialization/deserialization of objects.
 
     TODO: Three LRU cache levels are controlled, for device, host and disk.

dask_cuda/tests/test_dynamic_host_file.py → dask_cuda/tests/test_object_spilling_host_file.py

@@ -4,15 +4,15 @@
 from dask.dataframe.shuffle import shuffle_group
 from distributed import Client
 import dask_cuda
-from dask_cuda.dynamic_host_file import DynamicHostFile
+from dask_cuda.object_spilling_host_file import ObjectSpillingHostFile
 
 cupy = pytest.importorskip("cupy")
 cupy.cuda.set_allocator(None)
 itemsize = cupy.arange(1).nbytes
 
 
 def test_one_item_limit():
-    dhf = DynamicHostFile(device_memory_limit=itemsize)
+    dhf = ObjectSpillingHostFile(device_memory_limit=itemsize)
     dhf["k1"] = cupy.arange(1) + 1
     dhf["k2"] = cupy.arange(1) + 2
 
@@ -48,15 +48,15 @@ def test_one_item_limit():
     assert not dhf["k3"][1]._obj_pxy_serialized()
 
 
-@pytest.mark.parametrize("dynamic_spill", [True, False])
-def test_local_cuda_cluster(dynamic_spill):
+@pytest.mark.parametrize("object_spilling", [True, False])
+def test_local_cuda_cluster(object_spilling):
     """Testing spilling of a proxied cudf dataframe in a local cuda cluster"""
     cudf = pytest.importorskip("cudf")
     dask_cudf = pytest.importorskip("dask_cudf")
 
     def task(x):
         assert isinstance(x, cudf.DataFrame)
-        if dynamic_spill:
+        if object_spilling:
             # Check that `x` is a proxy object and the proxied DataFrame is serialized
             assert type(x) is dask_cuda.proxy_object.ProxyObject
             assert x._obj_pxy_get_meta()["serializers"] == ["dask", "pickle"]
@@ -67,7 +67,7 @@ def task(x):
 
     # Notice, setting `device_memory_limit=1B` to trigger spilling
     with dask_cuda.LocalCUDACluster(
-        n_workers=1, device_memory_limit="1B", dynamic_spill=dynamic_spill
+        n_workers=1, device_memory_limit="1B", object_spilling=object_spilling
     ) as cluster:
         with Client(cluster):
             df = cudf.DataFrame({"a": range(10)})
@@ -89,7 +89,7 @@ def test_dataframes_share_dev_mem():
     # They still share the same underlying device memory
     assert view1["a"].data._owner._owner is view2["a"].data._owner._owner
 
-    dhf = DynamicHostFile(device_memory_limit=160)
+    dhf = ObjectSpillingHostFile(device_memory_limit=160)
     dhf["v1"] = view1
     dhf["v2"] = view2
     v1 = dhf["v1"]