public class DeviceMemoryBuffer extends BaseDeviceMemoryBuffer
MemoryBuffer.EventHandler, MemoryBuffer.MemoryBufferCleaner| 修饰符和类型 | 方法和描述 |
|---|---|
static DeviceMemoryBuffer |
allocate(long bytes)
分配用于 GPU 的内存。
|
static DeviceMemoryBuffer |
allocate(long bytes, Cuda.Stream stream)
分配用于 GPU 的内存。
|
static DeviceMemoryBuffer |
fromRmm(long address, long lengthInBytes, long rmmBufferAddress)
将现有的 RMM 分配包装到设备内存缓冲区中。
|
DeviceMemoryBuffer |
slice(long offset, long len)
切分设备缓冲区的一部分。
|
copyFromDeviceBufferAsync, copyFromHostBuffer, copyFromHostBuffer, copyFromHostBuffer, copyFromHostBuffer, copyFromHostBuffer, copyFromHostBufferAsync, copyFromHostBufferAsync, sliceWithCopyaddressOutOfBoundsCheck, close, copyFromMemoryBuffer, copyFromMemoryBufferAsync, getAddress, getEventHandler, getLength, getRefCount, incRefCount, noWarnLeakExpected, setEventHandler, toStringpublic static DeviceMemoryBuffer fromRmm(long address, long lengthInBytes, long rmmBufferAddress)
address - RMM 分配的设备地址lengthInBytes - RMM 分配的长度，单位为字节rmmBufferAddress - 拥有设备内存的 rmm::device_buffer 的主机地址public static DeviceMemoryBuffer allocate(long bytes)
bytes - 要分配的大小，单位为字节public static DeviceMemoryBuffer allocate(long bytes, Cuda.Stream stream)
bytes - 要分配的大小，单位为字节stream - 用于同步此命令的流public final DeviceMemoryBuffer slice(long offset, long len)
slice 在类中 MemoryBufferoffset - 切片的起始位置。len - 要切分的字节数Copyright © 2025. 保留所有权利。