eupolicy.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
This Mastodon server is a friendly and respectful discussion space for people working in areas related to EU policy. When you request to create an account, please tell us something about you.

Server stats:

225
active users

#lmdb

5 posts3 participants0 posts today

Investigate #LMDB as HLR storage
projects.osmocom.org/issues/36

This Open Source Mobile Comm project appears to be an open source impl of a 3G network's backend software. But they're not using the standard 3GPP protocols, they're using their own home-grown stuff instead.

Looks like a toy, using SQLite3 as its HLR storage. Real 3G/4G networks use LDAP, lots of OpenLDAP in particular because nothing else scales enough, efficiently enough.

projects.osmocom.orgFeature #3649: Investigate LMDB as HLR data storage - OsmoHLR - Open Source Mobile CommunicationsRedmine

Evaluating Optimization Strategies for Dataset Storage for Machine Learning Workloads on HPC Systems

Bachelor Thesis

itc.rwth-aachen.de/cms/it-cent

Funny that their "example B+Tree structure as used by #LMDB" is not at all how LMDB is structured. It's not possible for a leaf page to provide a single pointer to its sibling page when using MVCC. The original LMDB slides illustrate the actual structure. Seems they didn't actually read the paper.

A fast CU partitioning algorithm for VVC based on dual-branch prediction network and hybrid mode decision
sciencedirect.com/science/arti

The latest Versatile Video Coding standard (VVC/H.266) introduces a Quad-Tree (QT) plus Nested Multi-Type Tree (QT-MTT) partitioning structure, significantly improving coding efficiency but resulting in high time complexity.

The #lmdb [57] library accelerates data loading during model training and achieves efficient data retrieval.

RiNALMo: general-purpose RNA language models can generalize well on structure prediction tasks

Motivated by the successes of protein language models, we introduce RiboNucleic Acid Language Model (RiNALMo) to unveil the hidden code of RNA. RiNALMo is the largest RNA language model to date, with 650M parameters pre-trained on 36M non-coding RNA sequences from several databases.

nature.com/articles/s41467-025

Built on #LMDB

NatureRiNALMo: general-purpose RNA language models can generalize well on structure prediction tasks - Nature CommunicationsRiNALMo, a large-scale RNA language model trained on non-coding RNA sequences, captures structural information and achieves state-of-the-art performance on multiple tasks, notably generalizing to unseen RNA families in secondary structure prediction.

Executor 2000

This is a modern fork of Executor (as of 2019)
Why "2000"? Because that's what we used to call the future.

Executor is a commercial Mac emulator from the 90s - what makes it different from other Mac emulators is that it doesn't require ROM files or any original Apple software to run; instead, it tries to re-implement the classic Mac OS API, just like WINE does for Windows.

The emulator's filesystem is built on #LMDB gitcode.com/gh_mirrors/executo

favicon
gitcode.comGitCode - 全球开发者的开源社区,开源代码托管平台GitCode是面向全球开发者的开源社区,包括原创博客,开源代码托管,代码协作,项目管理等。与开发者社区互动,提升您的研发效率和质量。

A Benchmark for Quantum Chemistry Relaxations via Machine Learning Interatomic Potentials

arxiv.org/html/2506.23008v1

"To efficiently store and access the parsed data, we save all the parsed trajectories into Lightning Memory-Mapped Database (#LMDB) files, a high-performance key–value storage format well-suited for large-scale machine learning pipelines."

arxiv.orgA Benchmark for Quantum Chemistry Relaxations via Machine Learning Interatomic Potentials

Sindre
A Python tool library that integrates multiple functions such as

🔍 #LMDB database management: high-performance key-value storage, support for multi-process operations
🎯 3D data processing: mesh processing, point cloud operations, 3D visualization
📊 Report generation: automatic generation of HTML test reports
🚀 Deployment tools: deployment solutions such as ONNX, TensorRT, shared memory, etc.

pypi.org/project/sindre/

pypi.orgClient Challenge

This guy on Linkedin just discovered #LMDB - I think he's in for a treat. linkedin.com/posts/pranjal-sha

How 'bout you, how/when did you first learn about it?

www.linkedin.com🧠 Just learned about LMDB today — and wow, it’s fascinating! | Pranjal Sharma🧠 Just learned about LMDB today — and wow, it’s fascinating! I was exploring database internals and came across LMDB (Lightning Memory-Mapped Database) — a tool that seems incredibly powerful but surprisingly under the radar. I haven’t used it in a project yet, but here’s what I’ve understood so far: 🔹 What is it? LMDB is an embedded key-value store that maps your entire database file directly into memory. This allows your app to read data as if it’s accessing RAM — super fast, zero-copy reads. 🔹 How does it work? Uses memory-mapped files for efficiency Data is stored in a B+ tree structure Copy-on-write keeps data consistent during writes ACID-compliant, with no background threads or WAL Allows multiple readers at the same time, without locking Only one writer at a time (a trade-off for simplicity + reliability) 🔹 Where is it used? Common in: Embedded systems Blockchain clients AI/ML workloads Anywhere you need fast, local, reliable storage To me, it feels like SQLite for key-value storage — lightweight, fast, and rock solid. 📚 I’ll definitely be digging deeper, maybe try it in a project soon. Just thought I’d share in case others haven’t come across it yet — it’s a hidden gem worth knowing about! Have you used LMDB? Curious to hear your thoughts or real-world use cases.

Just discovered this Japanese language group "London Tech Talk". Their book club has been reviewing Database Internals

london-tech-talk.com/episodes/

There's some mention of #LMDB but frankly very little. Kinda makes sense, all of the core ideas used in LMDB have been around since the 1960s or 70s. I.e. not much new, and all the modern buzz is around LSMs, etc.

But that's also why LMDB is bulletproof - all the techniques are time tested and well understood, while LSM tuning is still poke'n'pray.

London Tech Talk164. 【Bookclub 第四弾】 "Database Internals" #6 振り返り | London Tech Talk<p>London Tech Talk 名物 Bookclub 第四弾 "Database Internals" 第六章の振り返り収録です。"B-Tree Varients" の内容について振り返りました。</p><p>まず B Tree (B+ Tree) の簡単なおさらいをしました。B Tree のかかえる Write Amplification や Space Amplification という課題を解決するために B Tree の亜種が開発される必要があったという前提についてまずは理解しましょう。</p><p>最初に紹介したのは Lazy B-Trees です。Buffering というテクニックを利用して書き込みによる木構造の Node Split や Node Merge の頻度を下げる工夫について紹介しました。具体的には MongoDB のストレージエンジンでもある WiredTiger や Lazy-Adaptive Tree について触れました。</p><p>続いて Flash-Disk Tree (FD-Tree) について紹介しました。Buffering の議論を推し進めて、Chapter 7. でも紹介する LSM Trees をインスパイアーしたとされる FD Tree のキーワードについて紹介しました。</p><p>最後に、Microsoft から 2013 年に発表された Buzzword-Tree について紹介しました。マルチコア環境でもスケールするためのデータ構造として、CPU の特別な命令セットである CAS 命令を設計のコアに据え、Mapping Table と呼ばれる中間キャッシュレイヤーを実装した Buzzword-Tree の革新的なアーキテクチャについて触れました。</p><p>そのほか Bookclub で盛り上がった観点や、次回の Chapter 7 の予定について触れました。</p><p>ご意見・ご感想など、お便りはこちらの<a href="https://forms.gle/xFVHRRNehh9xFgKT8" target="_blank" rel="noopener noreferrer">⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠Google Form⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠</a> で募集しています。</p>