类使用
ai.rapids.cudf.BaseDeviceMemoryBuffer

使用 BaseDeviceMemoryBuffer 的程序包

程序包	描述
ai.rapids.cudf
ai.rapids.cudf.nvcomp

BaseDeviceMemoryBuffer 在 ai.rapids.cudf 中的使用

BaseDeviceMemoryBuffer 在 ai.rapids.cudf 中的子类

修饰符和类型	类和描述
类	CudaMemoryBuffer 此类表示直接使用 `cudaMalloc` 而不是默认 RMM 内存资源分配的数据。
类	CuFileBuffer 表示一个 cuFile 缓冲区。
类	DeviceMemoryBuffer 此类表示 GPU 上某种形式的数据。
类	DeviceMemoryBufferView 此类表示 GPU 上某种形式的数据。

ai.rapids.cudf 中返回 BaseDeviceMemoryBuffer 的方法

修饰符和类型	方法和描述
BaseDeviceMemoryBuffer	ColumnView.getData() 获取当前列视图 (viewHandle) 的数据缓冲区。
BaseDeviceMemoryBuffer	ColumnVector.OffHeapState.getData()
BaseDeviceMemoryBuffer	ColumnVector.getDeviceBufferFor(BufferType type) 获取此列的原始设备缓冲区访问权限。
BaseDeviceMemoryBuffer	ColumnView.getOffsets()
BaseDeviceMemoryBuffer	ColumnVector.OffHeapState.getOffsets()
BaseDeviceMemoryBuffer	ColumnView.getValid()
BaseDeviceMemoryBuffer	ColumnVector.OffHeapState.getValid()

ai.rapids.cudf 中参数类型为 BaseDeviceMemoryBuffer 的方法

修饰符和类型	方法和描述
static long	CuFile.appendDeviceBufferToFile(File path, BaseDeviceMemoryBuffer buffer) 将设备缓冲区同步追加到给定的文件路径。
void	HostMemoryBuffer.copyFromDeviceBuffer(BaseDeviceMemoryBuffer deviceMemoryBuffer) 将 DeviceMemoryBuffer 同步复制到 HostMemoryBuffer
void	HostMemoryBuffer.copyFromDeviceBuffer(BaseDeviceMemoryBuffer deviceMemoryBuffer, Cuda.Stream stream) 使用指定的流将 DeviceMemoryBuffer 复制到 HostMemoryBuffer。
void	HostMemoryBuffer.copyFromDeviceBufferAsync(BaseDeviceMemoryBuffer deviceMemoryBuffer, Cuda.Stream stream) 使用指定的流将 DeviceMemoryBuffer 复制到 HostMemoryBuffer。
void	BaseDeviceMemoryBuffer.copyFromDeviceBufferAsync(long destOffset, BaseDeviceMemoryBuffer src, long srcOffset, long length, Cuda.Stream stream) 使用指定的 CUDA 流将 src 的子集复制到此缓冲区，从 destOffset 开始。
static ColumnView	ColumnView.fromDeviceBuffer(BaseDeviceMemoryBuffer buffer, long startOffset, DType type, int rows) 从原始设备缓冲区创建新的列视图。
static void	CuFile.readFileToDeviceBuffer(BaseDeviceMemoryBuffer buffer, File path, long fileOffset) 将文件同步读入设备缓冲区。
static void	CuFile.writeDeviceBufferToFile(File path, long file_offset, BaseDeviceMemoryBuffer buffer) 将设备缓冲区同步写入给定的文件路径。

ai.rapids.cudf 中参数类型为 BaseDeviceMemoryBuffer 的构造器

构造器和描述
ColumnView(DType type, long rows, Optional<Long> nullCount, BaseDeviceMemoryBuffer dataBuffer, BaseDeviceMemoryBuffer validityBuffer) 根据设备上已有的数据创建新的列视图。
ColumnView(DType type, long rows, Optional<Long> nullCount, BaseDeviceMemoryBuffer dataBuffer, BaseDeviceMemoryBuffer validityBuffer, BaseDeviceMemoryBuffer offsetBuffer) 根据设备上已有的数据创建新的列视图。
ColumnView(DType type, long rows, Optional<Long> nullCount, BaseDeviceMemoryBuffer validityBuffer, BaseDeviceMemoryBuffer offsetBuffer, ColumnView[] children) 根据设备上已有的数据创建新的列视图。

BaseDeviceMemoryBuffer 在 ai.rapids.cudf.nvcomp 中的使用

ai.rapids.cudf.nvcomp 中参数类型为 BaseDeviceMemoryBuffer 的方法

修饰符和类型	方法和描述
DeviceMemoryBuffer[]	BatchedCompressor.compress(BaseDeviceMemoryBuffer[] origInputs, Cuda.Stream stream) 压缩一批缓冲区。
void	BatchedDecompressor.decompressAsync(BaseDeviceMemoryBuffer[] origInputs, BaseDeviceMemoryBuffer[] outputs, Cuda.Stream stream) 异步解压缩一批缓冲区
void	BatchedDecompressor.decompressAsync(BaseDeviceMemoryBuffer[] origInputs, BaseDeviceMemoryBuffer[] outputs, Cuda.Stream stream) 异步解压缩一批缓冲区
static void	BatchedLZ4Decompressor.decompressAsync(long chunkSize, BaseDeviceMemoryBuffer[] origInputs, BaseDeviceMemoryBuffer[] outputs, Cuda.Stream stream) 异步解压缩一批缓冲区
static void	BatchedLZ4Decompressor.decompressAsync(long chunkSize, BaseDeviceMemoryBuffer[] origInputs, BaseDeviceMemoryBuffer[] outputs, Cuda.Stream stream) 异步解压缩一批缓冲区