Module memory 

Memory Management Infrastructure

Provides GPU memory pooling, pressure handling, and analytics context management for high-performance kernel execution.

Features

• Memory Pools: Size-stratified pools for efficient GPU allocation
• Pressure Handling: Automatic memory pressure detection and mitigation
• Analytics Context: Reusable buffers for analytics workloads
• Reduction Buffers: Cached buffers for GPU reduction operations
• Multi-phase Reductions: Synchronization primitives for iterative algorithms

Example

use rustkernel_core::memory::{KernelMemoryManager, MemoryConfig};

let config = MemoryConfig::production();
let manager = KernelMemoryManager::new(config);

// Allocate from pool
let buffer = manager.allocate(1024 * 1024)?; // 1MB

// Return to pool
manager.deallocate(buffer);

Re-exports

pub use reduction::CooperativeBarrier;

pub use reduction::GlobalReduction;

pub use reduction::InterPhaseReduction;

pub use reduction::PhaseState;

pub use reduction::ReductionBuilder;

pub use reduction::ReductionConfig;

pub use reduction::ReductionError;

pub use reduction::ReductionOp;

pub use reduction::SyncMode;

Modules

reduction

Reduction Primitives

Structs

AnalyticsContext

Analytics context for reusable buffers

AnalyticsContextManager

Analytics context manager

BucketStats

Statistics for a size bucket

KernelMemoryManager

Kernel memory manager

MemoryBuffer

Memory buffer handle

MemoryConfig

Memory configuration

MemoryStats

Memory statistics

ReductionBuffer

Reduction buffer for GPU reduction operations

ReductionBufferCache

Reduction buffer cache

SizeBucket

Size bucket for memory pool

Enums

MemoryError

Memory errors

PressureLevel

Memory pressure level

Type Aliases

AllocResult

Memory allocation result