scirs2-core 0.4.2

//! Memory-efficient operations and views for large arrays.
//!
//! This module provides utilities for working with large arrays efficiently by:
//! - Using chunk-wise processing to reduce memory requirements
//! - Creating memory-efficient views that avoid full data copies
//! - Implementing lazy evaluation and operation fusion for improved performance
//! - Supporting out-of-core processing for data that doesn't fit in RAM
//! - Smart prefetching for improved performance with predictable access patterns

mod adaptive_chunking;
mod adaptive_feedback;
mod adaptive_prefetch;
/// Async out-of-core I/O with LRU caching and chunk-based streaming
pub mod async_out_of_core;
mod chunk_format;
mod chunked;
#[cfg(feature = "memory_compression")]
mod compressed_memmap;
#[cfg(feature = "gpu")]
mod cross_device;
mod cross_file_prefetch;
mod fusion;
mod large_data;
mod lazy_array;
mod memmap;
mod memmap_chunks;
mod memmap_slice;
mod memory_layout;
/// Memory-mapped array wrapper with COW semantics and zero-copy views
pub mod mmap_array;
mod numa_topology;
mod out_of_core;
mod out_of_core_v2;
mod pattern_recognition;
mod platform_memory;
mod prefetch;
mod resource_aware;
#[cfg(feature = "parallel")]
mod streaming;
mod validation;
mod views;
#[cfg(feature = "parallel")]
mod work_stealing;
mod zero_copy_interface;
#[cfg(feature = "parallel")]
mod zero_copy_streaming;
mod zero_serialization;
mod zerocopy;

pub use adaptive_chunking::{
    create_shared_monitor,
    // Core types
    AdaptiveChunking,
    AdaptiveChunkingBuilder,
    AdaptiveChunkingParams,
    AdaptiveChunkingResult,
    // v0.2.0: Memory management types
    AdaptiveChunkingWithMonitoring,
    ChunkProcessingStats,
    DynamicChunkAdjuster,
    MemoryLimits,
    MemoryPressureLevel,
    MemoryPressureMonitor,
    MemoryTrend,
    SharedMemoryMonitor,
    WorkloadType,
};
pub use adaptive_feedback::{
    create_shared_predictor, ChunkSizePredictor, PerformanceMetrics, SharedPredictor,
};
pub use adaptive_prefetch::{
    AdaptivePatternTracker, AdaptivePrefetchConfig, AdaptivePrefetchConfigBuilder,
    PatternTrackerFactory, PrefetchStrategy,
};
pub use chunk_format::{
    detect_format_version, read_chunk_index, read_header, write_chunk_index, write_header,
    ChunkIndex, ChunkIndexEntry, CompressionType, FormatVersion, OutOfCoreHeaderV2,
    HEADER_FIXED_SIZE, MAGIC_BYTES_V2,
};
pub use chunked::{
    chunk_wise_binary_op, chunk_wise_op, chunk_wise_reduce, ChunkedArray, ChunkingStrategy,
    OPTIMAL_CHUNK_SIZE,
};
#[cfg(feature = "memory_compression")]
pub use compressed_memmap::{
    CompressedFileMetadata, CompressedMemMapBuilder, CompressedMemMappedArray, CompressionAlgorithm,
};
#[cfg(feature = "gpu")]
pub use cross_device::{
    create_cpuarray, create_cross_device_manager, create_gpuarray, to_best_device,
    CrossDeviceManager, DeviceArray, DeviceBuffer, DeviceMemoryManager, DeviceMemoryPool,
    DeviceStream, DeviceType, KernelParam, ToDevice, ToHost, TransferDirection, TransferEvent,
    TransferMode, TransferOptions, TransferOptionsBuilder,
};
pub use cross_file_prefetch::{
    AccessType, CrossFilePrefetchConfig, CrossFilePrefetchConfigBuilder, CrossFilePrefetchManager,
    CrossFilePrefetchRegistry, DataAccess, DatasetId, DatasetPrefetcher,
};
pub use fusion::{register_fusion, FusedOp, OpFusion};
#[cfg(feature = "parallel")]
pub use large_data::{create_parallel_processor, ParallelStreamingProcessor};
pub use large_data::{create_streaming_iterator, StreamingChunkIterator};
pub use lazy_array::{evaluate, LazyArray, LazyOp, LazyOpKind};
pub use memmap::{create_mmap, create_temp_mmap, open_mmap, AccessMode, MemoryMappedArray};
#[cfg(feature = "parallel")]
pub use memmap_chunks::MemoryMappedChunksParallel;
pub use memmap_chunks::{ChunkIter, MemoryMappedChunkIter, MemoryMappedChunks};
pub use memmap_slice::{MemoryMappedSlice, MemoryMappedSlicing};
pub use memory_layout::{
    AccessPattern as MemoryAccessPattern, ArrayCreation, ArrayLayout, LayoutConverter, LayoutOrder,
    MemoryLayout,
};
pub use numa_topology::{NumaNode, NumaPolicy, NumaTopology};
pub use out_of_core::{create_disk_array, load_chunks, DiskBackedArray, OutOfCoreArray};
pub use out_of_core_v2::OutOfCoreArrayV2;
pub use pattern_recognition::{
    ComplexPattern, Confidence, PatternRecognitionConfig, PatternRecognizer, RecognizedPattern,
};
pub use platform_memory::PlatformMemoryInfo;
#[cfg(feature = "memory_compression")]
pub use prefetch::PrefetchingCompressedArray;
pub use prefetch::{
    AccessPattern, PrefetchConfig, PrefetchConfigBuilder, PrefetchStats, Prefetching,
};
pub use resource_aware::{
    ResourceAwareConfig, ResourceAwareConfigBuilder, ResourceAwarePrefetcher, ResourceMonitor,
    ResourceSnapshot, ResourceSummary, ResourceType,
};
#[cfg(feature = "parallel")]
pub use streaming::{
    create_pipeline, create_stream_processor, Pipeline, PipelineBuilder, PipelineStats,
    StreamConfig, StreamConfigBuilder, StreamMode, StreamProcessor, StreamSource, StreamState,
    StreamStats,
};
pub use views::{diagonal_view, transpose_view, view_as, view_mut_as, ArrayView, ViewMut};
#[cfg(feature = "parallel")]
pub use work_stealing::{
    create_cpu_intensive_scheduler, create_io_intensive_scheduler, create_work_stealing_scheduler,
    SchedulerStats, TaskPriority, WorkStealingConfig, WorkStealingConfigBuilder,
    WorkStealingScheduler, WorkStealingTask,
};
pub use zero_copy_interface::{
    create_global_data_registry, create_zero_copy_data, get_global_data, global_interface,
    register_global_data, DataExchange, DataId, DataMetadata, FromZeroCopy, InterfaceStats,
    IntoZeroCopy, ZeroCopyData, ZeroCopyInterface, ZeroCopyView, ZeroCopyWeakRef,
};
#[cfg(feature = "parallel")]
pub use zero_copy_streaming::{
    create_zero_copy_processor, BufferPool, BufferPoolStats, LockFreeQueue, ProcessingMode,
    WorkStealingScheduler as ZeroCopyWorkStealingScheduler,
    WorkStealingTask as ZeroCopyWorkStealingTask, ZeroCopyBuffer, ZeroCopyConfig, ZeroCopyStats,
    ZeroCopyStreamProcessor,
};
pub use zero_serialization::{ZeroCopySerializable, ZeroCopySerialization};
pub use zerocopy::{ArithmeticOps, BroadcastOps, ZeroCopyOps};

// Re-export commonly used items in a prelude module for convenience
pub mod prelude {
    // Core functionality always available
    pub use super::{
        chunk_wise_binary_op, chunk_wise_op, chunk_wise_reduce, create_mmap, create_temp_mmap,
        evaluate, open_mmap, view_as, view_mut_as, AccessMode, AdaptiveChunking,
        AdaptiveChunkingBuilder, AdaptivePatternTracker, AdaptivePrefetchConfig, ArithmeticOps,
        ArrayCreation, ArrayLayout, ArrayView, BroadcastOps, ChunkIter, ChunkedArray,
        ComplexPattern, CrossFilePrefetchManager, DatasetId, DatasetPrefetcher, LayoutOrder,
        LazyArray, MemoryLayout, MemoryMappedArray, MemoryMappedChunkIter, MemoryMappedChunks,
        MemoryMappedSlice, MemoryMappedSlicing, OutOfCoreArray, PatternRecognizer, PrefetchConfig,
        PrefetchConfigBuilder, PrefetchStrategy, Prefetching, ResourceAwareConfig,
        ResourceAwarePrefetcher, ViewMut, ZeroCopyData, ZeroCopyInterface, ZeroCopyOps,
        ZeroCopySerializable, ZeroCopySerialization, ZeroCopyView,
    };

    // GPU-specific exports
    #[cfg(feature = "gpu")]
    pub use super::{
        create_cpuarray, create_cross_device_manager, create_gpuarray, to_best_device,
    };

    // Parallel processing exports
    #[cfg(feature = "parallel")]
    pub use super::MemoryMappedChunksParallel;
}