memory-optimization

This free RAM cleaner uses native Windows features to optimize memory areas. It's a compact, portable, and smart application.

Detect and fix suboptimal struct field layout in Go: reduces padding and GC pointer bytes, rewrites files while preserving comments. A drop-in fieldalignment replacement.

Run larger LLMs with longer contexts on Apple Silicon by using differentiated precision for KV cache quantization. KVSplit enables 8-bit keys & 4-bit values, reducing memory by 59% with <1% quality loss. Includes benchmarking, visualization, and one-command setup. Optimized for M1/M2/M3 Macs with Metal support.

A native macOS system toolkit: honest memory optimization, deep cleanup that finds the real space hogs, fan control, keep-awake, disk analyzer.

VS Code extension: Go struct layout, padding, reorder

A fast and memory-optimized string library for heavy-text manipulation in Python

Keep ChatGPT fast — Firefox & Chrome extension that trims DOM in long conversations

Materials about memory optimization and zero-allocation samples.

Flash weight streaming for MLX: run massive models larger than your RAM on Apple Silicon.

Free up your memory for free!

This code repository contains the code used for my "Optimizing Memory Usage for Training LLMs and Vision Transformers in PyTorch" blog post.

First open-source implementation of Google TurboQuant (ICLR 2026) -- near-optimal KV cache compression for LLM inference. 5x compression with near-zero quality loss.

Lightweight Python lazy imports that defer module loading to reduce startup time and initial memory use.

A performant and memory efficient storage for immutable strings with C++17. Supports all standard char types: char, wchar_t, char16_t, char32_t and C++20's char8_t.

Simple RAM Cleaner

🚀 Zero-config automatic memory allocator for Rust - just add one line and get up to 1.6x faster allocation performance across all platforms

A curated list of awesome optimizations that you can do to improve your redis deployment (both client side and server side).

Unified KV-cache compression for LLM inference: 12 Python-native methods, guarded add-on composition and routing, analytical capacity simulation, Godzilla KVarN/TriAttention with SM86/SM89 qualification, CUDA weight sharing, and multi-GPU planning.

PRISM: O(1) Photonic Block Selection for Long-Context LLM Inference — eliminates the O(N) KV cache scan via photonic broadcast-and-weight similarity engine on TFLN

A tiered-memory system design for workloads that don't fit in RAM: measure the working set, pin the hot tier, stream the cold tier from flash. Ships the residency calculator, measurement harnesses, and the build recipes behind it. Predictions validated against public benchmarks.