#cache locality
30 related articles
·3 min
Object Identity in Software: Core Principles of Interface, ECS, and Component Design
Exploring the core principle of separating object identity from representation in software design, covering interfaces, ECS, DDD, and distributed systems.
Read more →
·3 min
Training a 1.3B Parameter LLM from Scratch: Complete Workflow and Core Challenges
A deep dive into the complete workflow of training a 1.3B parameter LLM from scratch, covering Transformer architecture design, data preparation, and distributed training optimization.
Read more →
·3 min
Go Team Proposes Generic Collection Library 'container': A Deep Dive into the Standard Library Completion Plan
The Go team proposes new generic collection types under container/, including Set, ordered Map, queues, and more. A deep dive into the proposal and its design trade-offs.
Read more →
·4 min
Colibri: Running a 744-Billion-Parameter LLM on an Ordinary Laptop
Colibri uses MoE hot-cold separation and 4-bit quantization to run 744B-parameter models like GLM 5.2 on consumer hardware. Learn about its three-tier memory architecture and speculative decoding.
Read more →
·4 min
Building a High-Performance Lock-Free Queue from Scratch: A Modern C++ Practical Guide
A deep dive into building high-performance lock-free queues with modern C++, covering CAS atomic operations, memory ordering, Michael-Scott algorithm, ring buffers, ABA problem solutions, and memory reclamation mechanisms.
Read more →
·7 min
DeepSeek's Mysteriously Retracted Paper: Breaking AI's Multimodal Reasoning Bottleneck with Visual Primitives
DeepSeek's paper 'Thinking with Visual Primitives' was online for just 4 hours before being pulled. It uses bounding boxes and points as reasoning primitives, letting models 'point at' images to outperform GPT, Gemini, and Claude on maze navigation and counting.
Read more →
Can LLMs Really Understand Computer Ar…
·3 min
Can LLMs Really Understand Computer Architecture Papers? A Deep Dive into Comprehension Capabilities
Can LLMs truly understand computer architecture papers? This article analyzes core challenges—from surface pattern matching to deep reasoning—and defines their capability limits for researchers.
Read more →
·3 min
Bernini Tile-Based Generation: An Engineered Approach to Video Super-Resolution in ComfyUI
Bernini is a ComfyUI video super-resolution node package using Tile Split/Select/Merge to solve seams, drift, and VRAM overflow. Benchmarked at 325s for 39 frames at 1920×1080.
Read more →
Why Tokio and Rayon Don't Mix Well: A …
·2 min
Why Tokio and Rayon Don't Mix Well: A Deep Dive into Rust's Async Blocking Pitfalls
Why mixing Tokio and Rayon in Rust async code causes performance traps, how spawn_blocking isolates blocking tasks, and the key difference between concurrency and parallelism.
Read more →
·3 min
New Retrieval Paradigms: Hobbit's Hard Negative Construction and Disco's Collaborative Coverage Index Design
A deep dive into two frontier dense retrieval works: Hobbit uses gradient analysis to automatically construct hard batches; Disco replaces single-document competition with submodular collaborative coverage, reshaping Top-K retrieval.
Read more →
CUDA Kernel Fusion: A Practical Guide …
·2 min
CUDA Kernel Fusion: A Practical Guide to Reducing Memory Bandwidth and Launch Overhead
Learn how CUDA Kernel Fusion merges multiple GPU kernels to reduce global memory traffic and launch overhead, with real-world examples from AI inference and deep learning.
Read more →
Flash-MSA: How Sparse Attention Kernel…
·2 min
Flash-MSA: How Sparse Attention Kernels Break the Million-Token Training Barrier
Flash-MSA uses GPU sparse attention kernels to reduce complexity from O(n²) to near-linear, tackling the compute bottleneck of million-token LLM training.
Read more →
·3 min
Programs as Data Flow: Rethinking Architecture Paradigms Through Building a 3D Editor in C++
A developer builds a 3D editor from scratch in C++, modeling programs as dataflow graphs rather than object collections. Exploring dataflow vs. OOP, Greenspun's Tenth Rule, and the debugging dividend of visual programming.
Read more →
·3 min
CPU-Native Language Models: Breaking Through Inference Bottlenecks with Memory Bandwidth Co-Design
SiliconLLM builds a CPU-native LLM architecture from scratch, combining selective SSM, ternary (1.58-bit) LUT MLP, and granular MoE, co-designed around the L3 cache bandwidth cliff. Ternary kernels show 4-5x speedup over fp32.
Read more →
·3 min
Designing MPMC Queues: Balancing Bounded Waiting and High-Performance Concurrency
An in-depth look at the core design challenges of MPMC queues: bounded waiting, lock-free mechanisms, sequence-number slots, dual-path strategies, and helping. Helping engineers balance throughput, fairness, and latency.
Read more →
·4 min
Why Do AI Agents Crash the Moment They Go Live? A Practical Guide to Harness Engineering for Runtime
A deep dive into the Claude Code source code, systematically analyzing the five-layer Harness Engineering architecture: environment, tool, control, memory, and evaluation. Build a stable runtime system for production AI Agents.
Read more →
·4 min
meshoptimizer: A Deep Dive into This Open-Source 3D Mesh Compression and Rendering Optimization Library
meshoptimizer is a lightweight C++ open-source library offering vertex cache optimization, overdraw reduction, mesh compression, quantization, and LOD simplification for game engines, WebGL, and glTF pipelines.
Read more →
·3 min
Triple Dragon Fractal: Recursive Generation and the Beauty of Symmetry in Dragon Curves
Explore the math behind the Triple Dragon Fractal: from paperfolding dragon curves to L-systems, threefold symmetry, Hausdorff dimension, and fractal programming.
Read more →
·4 min
24 Hours, No AI, Zero to Rust: Building a Game from Scratch — A Full Account
A YouTuber spent 24 hours learning Rust from zero with no AI tools, mastering ownership, borrowing, and a game engine to build a complete Brick Breaker game. Full account of every challenge.
Read more →
·4 min
Ternlight: A 7MB Browser-Side Embedding Model for Server-Free Semantic Search
Ternlight is a 7MB WebAssembly-based browser-side text embedding model requiring no server or GPU. Explore its tech, use cases, and tradeoffs for private, offline semantic search.
Read more →