Package ai.rapids.cudf
Class RmmAllocationMode
java.lang.Object
ai.rapids.cudf.RmmAllocationMode
-
Field Summary
FieldsModifier and TypeFieldDescriptionstatic final int
Use arena suballocation strategystatic final int
Use CUDA async suballocation strategystatic final int
Use CUDA async suballocation strategy with fabric handles that are peer accessible with read-write accessstatic final int
Use cudaMalloc for allocationstatic final int
Use cudaMallocManaged rather than cudaMallocstatic final int
Use pool suballocation strategy -
Constructor Summary
Constructors -
Method Summary
-
Field Details
-
CUDA_DEFAULT
public static final int CUDA_DEFAULTUse cudaMalloc for allocation- See Also:
-
POOL
public static final int POOLUse pool suballocation strategy- See Also:
-
CUDA_MANAGED_MEMORY
public static final int CUDA_MANAGED_MEMORYUse cudaMallocManaged rather than cudaMalloc- See Also:
-
ARENA
public static final int ARENAUse arena suballocation strategy- See Also:
-
CUDA_ASYNC
public static final int CUDA_ASYNCUse CUDA async suballocation strategy- See Also:
-
CUDA_ASYNC_FABRIC
public static final int CUDA_ASYNC_FABRICUse CUDA async suballocation strategy with fabric handles that are peer accessible with read-write access- See Also:
-
-
Constructor Details
-
RmmAllocationMode
public RmmAllocationMode()
-