类用法
ai.rapids.cudf.HostMemoryBuffer

使用 HostMemoryBuffer 的包

包	描述
ai.rapids.cudf

HostMemoryBuffer 在 ai.rapids.cudf 中的用法

声明为 HostMemoryBuffer 类型的 ai.rapids.cudf 中的字段

修饰符和类型	字段和描述
HostMemoryBuffer	HostColumnVectorCore.OffHeapState.data
HostMemoryBuffer	HostColumnVectorCore.OffHeapState.offsets
HostMemoryBuffer	HostColumnVectorCore.OffHeapState.valid

返回 HostMemoryBuffer 类型的 ai.rapids.cudf 中的方法

修饰符和类型	方法和描述
HostMemoryBuffer	DefaultHostMemoryAllocator.allocate(long bytes)
static HostMemoryBuffer	PinnedMemoryPool.allocate(long bytes) 创建主机缓冲区的工厂方法，但优先指向固定内存。
static HostMemoryBuffer	HostMemoryBuffer.allocate(long bytes) 分配内存，但务必关闭返回的缓冲区以避免内存泄漏。
HostMemoryBuffer	HostMemoryAllocator.allocate(long bytes) 分配内存，但务必关闭返回的缓冲区以避免内存泄漏。
HostMemoryBuffer	DefaultHostMemoryAllocator.allocate(long bytes, boolean preferPinned)
static HostMemoryBuffer	HostMemoryBuffer.allocate(long bytes, boolean preferPinned) 分配内存，但务必关闭返回的缓冲区以避免内存泄漏。
HostMemoryBuffer	HostMemoryAllocator.allocate(long bytes, boolean preferPinned) 分配内存，但务必关闭返回的缓冲区以避免内存泄漏。
static HostMemoryBuffer	PinnedMemoryPool.allocate(long bytes, HostMemoryAllocator hostMemoryAllocator) 创建主机缓冲区的工厂方法，但优先指向固定内存。
static HostMemoryBuffer	HostMemoryBuffer.allocateRaw(long bytes) 绕过默认分配器分配主机内存。
HostMemoryBuffer	HostColumnVectorCore.getData() 返回给定主机端列向量的数据缓冲区
HostMemoryBuffer	JCudfSerialization.HostConcatResult.getHostBuffer()
HostMemoryBuffer	HostColumnVectorCore.getHostBufferFor(BufferType type) 获取此列的原始主机缓冲区的访问权限。
HostMemoryBuffer	HostColumnVectorCore.getOffsets() 返回偏移缓冲区
HostMemoryBuffer	HostColumnVectorCore.getValidity() 返回给定主机端列向量的有效性缓冲区
HostMemoryBuffer	MultiBufferDataSource.hostRead(long offset, long amount)
abstract HostMemoryBuffer	DataSource.hostRead(long offset, long amount) 从源的给定偏移量处读取数据。
static HostMemoryBuffer	HostMemoryBuffer.mapFile(File path, FileChannel.MapMode mode, long offset, long length) 创建映射到文件的主机缓冲区。
HostMemoryBuffer	HostMemoryBuffer.slice(long offset, long len) 切片主机缓冲区的一部分。
HostMemoryBuffer	HostMemoryBuffer.sliceWithCopy(long offset, long len) 切片主机缓冲区的一部分，并实际复制数据。
static HostMemoryBuffer	PinnedMemoryPool.tryAllocate(long bytes) 创建固定主机内存缓冲区的工厂方法。

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

修饰符和类型	方法和描述
static ContiguousTable	JCudfSerialization.concatToContiguousTable(JCudfSerialization.SerializedTableHeader[] headers, HostMemoryBuffer[] dataBuffers) 将主机内存中的多个表连接成设备内存中的连续表。
static JCudfSerialization.HostConcatResult	JCudfSerialization.concatToHostBuffer(JCudfSerialization.SerializedTableHeader[] headers, HostMemoryBuffer[] dataBuffers)
static JCudfSerialization.HostConcatResult	JCudfSerialization.concatToHostBuffer(JCudfSerialization.SerializedTableHeader[] headers, HostMemoryBuffer[] dataBuffers, HostMemoryAllocator hostMemoryAllocator) 将主机内存中的多个表连接成单个主机表缓冲区。
void	BaseDeviceMemoryBuffer.copyFromHostBuffer(HostMemoryBuffer src) 从 src 的开头复制所有内容到此缓冲区的开头。
void	BaseDeviceMemoryBuffer.copyFromHostBuffer(HostMemoryBuffer src, Cuda.Stream stream) 使用 CUDA 流从此缓冲区的开头复制整个主机缓冲区。
void	BaseDeviceMemoryBuffer.copyFromHostBuffer(HostMemoryBuffer src, long srcOffset, long length) 从 src 的子集复制到此缓冲区的开头。
void	BaseDeviceMemoryBuffer.copyFromHostBuffer(long destOffset, HostMemoryBuffer src, long srcOffset, long length) 从 src 的子集复制到此缓冲区的 destOffset 位置。
void	HostMemoryBuffer.copyFromHostBuffer(long destOffset, HostMemoryBuffer srcData, long srcOffset, long length) 将给定缓冲区的内容复制到此缓冲区
void	BaseDeviceMemoryBuffer.copyFromHostBuffer(long destOffset, HostMemoryBuffer src, long srcOffset, long length, Cuda.Stream stream) 使用指定的 CUDA 流从 src 的子集复制到此缓冲区的 destOffset 位置。
void	BaseDeviceMemoryBuffer.copyFromHostBufferAsync(HostMemoryBuffer src, Cuda.Stream stream) 使用 CUDA 流从此缓冲区的开头复制整个主机缓冲区。
void	BaseDeviceMemoryBuffer.copyFromHostBufferAsync(long destOffset, HostMemoryBuffer src, long srcOffset, long length, Cuda.Stream stream) 使用指定的 CUDA 流从 src 的子集复制到此缓冲区的 destOffset 位置。
void	HostBufferConsumer.handleBuffer(HostMemoryBuffer buffer, long len) 消费一个缓冲区。
long	MultiBufferDataSource.hostRead(long offset, HostMemoryBuffer dest)
abstract long	DataSource.hostRead(long offset, HostMemoryBuffer dest) 从源的给定偏移量处读取数据到 dest。
protected void	DataSource.onHostBufferDone(HostMemoryBuffer buffer) 当 hostRead 返回的缓冲区完成时调用。
static Table	JCudfSerialization.readAndConcat(JCudfSerialization.SerializedTableHeader[] headers, HostMemoryBuffer[] dataBuffers)
static Table	Table.readAvro(AvroOptions opts, HostMemoryBuffer buffer, long offset, long len) 读取 Avro 格式的数据。
static Table	Table.readCSV(Schema schema, CSVOptions opts, HostMemoryBuffer buffer, long offset, long len) 读取 CSV 格式的数据。
long	HostBufferProvider.readInto(HostMemoryBuffer buffer, long len) 将数据放入给定缓冲区。
static TableWithMeta	Table.readJSON(JSONOptions opts, HostMemoryBuffer buffer, long offset, long len) 读取 JSON 格式的数据并推断列名和 schema。
static Table	Table.readJSON(Schema schema, JSONOptions opts, HostMemoryBuffer buffer, long offset, long len) 读取 JSON 格式的数据。
static Table	Table.readJSON(Schema schema, JSONOptions opts, HostMemoryBuffer buffer, long offset, long len, int emptyRowCount) 已过时。 此方法已过时，因为 emptyRowCount 未使用。请改用不带 emptyRowCount 的方法。
static Table	Table.readORC(ORCOptions opts, HostMemoryBuffer buffer, long offset, long len) 读取 ORC 格式的数据。
static Table	Table.readParquet(ParquetOptions opts, HostMemoryBuffer... buffers) 读取 Parquet 格式的数据。
static Table	Table.readParquet(ParquetOptions opts, HostMemoryBuffer buffer, long offset, long len) 读取 Parquet 格式的数据。
static JCudfSerialization.TableAndRowCountPair	JCudfSerialization.readTableFrom(JCudfSerialization.SerializedTableHeader header, HostMemoryBuffer hostBuffer)
static void	JCudfSerialization.readTableIntoBuffer(InputStream in, JCudfSerialization.SerializedTableHeader header, HostMemoryBuffer buffer) 读取表的 header 后，将数据部分读取到主机端缓冲区中。
static HostColumnVector[]	JCudfSerialization.unpackHostColumnVectors(JCudfSerialization.SerializedTableHeader header, HostMemoryBuffer hostBuffer) 将序列化的连续表反序列化为主机列数组。
static void	JCudfSerialization.writeConcatedStream(JCudfSerialization.SerializedTableHeader[] headers, HostMemoryBuffer[] dataBuffers, OutputStream out) 从存储在已解析 headers 和 dataBuffer 中的多个批次中获取数据，并将其写入 out，就像它是一个单独的缓冲区一样。

ai.rapids.cudf 中参数类型为 HostMemoryBuffer 的构造方法

构造方法和描述
HostColumnVector(DType type, long rows, Optional<Long> nullCount, HostMemoryBuffer hostDataBuffer, HostMemoryBuffer hostValidityBuffer, HostMemoryBuffer offsetBuffer, List<HostColumnVectorCore> nestedHcv) 创建在主机上填充了数据的新列向量。
HostColumnVectorCore(DType type, long rows, Optional<Long> nullCount, HostMemoryBuffer data, HostMemoryBuffer validity, HostMemoryBuffer offsets, List<HostColumnVectorCore> nestedChildren)
HostConcatResult(JCudfSerialization.SerializedTableHeader tableHeader, HostMemoryBuffer tableBuffer)
MultiBufferDataSource(HostMemoryAllocator allocator, HostMemoryBuffer... buffers) 创建由多个缓冲区支持的新数据源。
MultiBufferDataSource(HostMemoryBuffer... buffers) 创建由多个缓冲区支持的新数据源。
ORCChunkedReader(long chunkReadLimit, long passReadLimit, long outputRowSizingGranularity, ORCOptions opts, HostMemoryBuffer buffer, long offset, long len) 构造分块 ORC 读取器实例，类似于 ORCChunkedReader(long, long, ORCOptions, HostMemoryBuffer, long, long)，增加了一个额外参数来控制输出表的粒度。
ORCChunkedReader(long chunkReadLimit, long passReadLimit, ORCOptions opts, HostMemoryBuffer buffer, long offset, long len) 从读取限制、输出行粒度以及已加载到内存缓冲区中的文件构造读取器实例。
ParquetChunkedReader(long chunkSizeByteLimit, long passReadLimit, ParquetOptions opts, HostMemoryBuffer... buffers) 从读取限制和主机内存缓冲区中的数据构造读取器实例。
ParquetChunkedReader(long chunkSizeByteLimit, long passReadLimit, ParquetOptions opts, HostMemoryBuffer buffer, long offset, long len) 从读取限制和已读取到内存缓冲区中的文件构造读取器实例。
ParquetChunkedReader(long chunkSizeByteLimit, ParquetOptions opts, HostMemoryBuffer buffer, long offset, long len) 从读取限制和已读取到内存缓冲区中的文件构造读取器实例。