Bogdan Melnik

Here are the slides for my GopherCon.EU talk: How to Win Frames and Influence Pointers https://lnkd.in/d7byRQfu #gopherconEU #golang

72

4 Comments

Another blow hits C and C++, considered memory-unsafe languages. US companies are requested to avoid them in new projects and define a memory-safety roadmap for existing code. What does this mean for EU and C-only coding shops? #memoryunsafe #clanguage #programming

18

9 Comments

The current lack of resources allocated for the development of technologies that support the Georgian language is a concern. This scarcity not only suppresses local innovation but also restricts the potential societal benefits that such tools could bring to our nation. Moreover, the reluctance to digitize and share textual data within Georgia worsens the issue, further limiting the availability of essential resources necessary for technological advancement and cultural preservation. This hesitation severely slows down progress in educational, technological, and cultural sectors.

6

2 Comments

Found this channel called "Programmers Are Also Human" with funny skids on many programming languages, editors, dev processes and software development roles in general. Pick your favorite language, editor, process or role in SW development and have a 😂 . This one is on Rust to get the idea. "An interpreter based language needs a lot of imagination" 😂 #rust #funny #jokes #programming #programmersarealsohuman #swfunny #swdeveloperhumor #humor

Roofline Analysis of Apple Silicon GPUs using MLX I wrote a tutorial (https://lnkd.in/d8N6_8aU) on Roofline analysis and its application to Apple silicon GPU kernel development, a method traditionally used by CUDA kernel developers. A big thank you to Angelos Katharopoulos and Awni Hannun, the creators of MLX, for their invaluable help in making this tutorial possible. The inspiration came from friends who asked why Recurrent Drafting achieves higher speedups on the M2 Ultra compared to the M1 Max after reading my previous post announcing the open-source of its MLX implementation (https://lnkd.in/dNtdfyzW). The Roofline model helps explain the performance differences by analyzing the computational and memory bandwidth limits of the hardware. I hope this tutorial shares my understanding about performance optimization techniques for Apple silicon GPUs and showcases the outstanding performance of MLX as well as the benefits it brings to applications like Recurrent Drafting. Your feedback is welcome!

120

5 Comments

So I became fascinated by C++ coroutines late last week, and have spent a good portion of the weekend figuring out how they work. The TL;DR of my discoveries so far is that C++20 coroutines: 1) follow the best and the worst of C++ practices, and 2) are a really clever way to decouple stackless functions from their execution model. For (1), well, it's C++. No crisp & clean syntax, lots of weird compilation errors, and even weirder runtime errors. I hated them with passion for the first several hours. But then they grew on me, because the flip side is also true. The flip side is backwards compatibility and "only pay for what you use". And introducing a totally new concept — the stackless execution model, effectively, [proto-] green threads — into a mature and complex language is one hell of a pain! For (2), I do believe it's a genius solution for the language to separate stackless functions from their execution model. And C++20 did exactly that. There is literally no event loop in C++20! The claim (as I understand it) is that a tiny microcontroller may want to use C++20 coroutines, and the authors of the Standard are very explicit about not constraining possible runtimes. All that C++20 coroutines do is introduce stackless functions into the language as first-class citizens. And yes, stackless immediately implies resumable. So it is one step about the most-clever-possible syntactic sugar over `setjmp()`. Full one step. But not a millimeter more than this one step. Whether you "execute" your coroutines in a dedicated single-threaded event loop, or a thread-pool-based event loop, or by individual steps via some timer event or even a network packet — that is entirely your call. The stackless functions are there, `co_await` (as well as `co_yield` and `co_return`) is there, no callback hell is needed, and all the RAII and scoped initialization/ownership model is fully intact. Now, I'm still elbows deep in my experiments, but here's the first big LinkedIn-worthy question. Why don't we write our C++20 [unit] tests as corourines? Say, you want to limit some test to five seconds. And fail it as "timed out" within these five seconds. Before coroutines, doing this without instrumenting the very code of the test involved `fork()`-ing the process, effectively relying on the operating system's primitives to ensure proper resource deallocation. With C++20 coroutines, as long as the slow test is slow on non-blocking I/O (i.e. not in an infinite loop, but on something that `co_await`-s stuff), it is possible to time-box each individual test to some five seconds very much organically. Because a coroutine frame, unlike a stack frame, can be `.destroy()`-ed safely and cleanly. So, are we doing it yet? If no, why not? If yes, what would be some next-gen `gtest` to look into? Also, I'll be writing some code and blog posts on C++20 coroutines soon. If you are reading this text and are an expert, DM me pls — I'd love to pick your brain!

19

4 Comments

𝐍𝐚𝐭𝐢𝐯𝐞 𝐇𝐓𝐓𝐏'𝐬 𝐢𝐧 𝐆𝐨𝐥𝐚𝐧𝐠! 💡 Another new and interesting article to dive into around Golang! https://lnkd.in/ea5gYBVt As always please leave any impressions and thoughts in the comments! 😁

22

1 Comment

My take on the 'russian kernel maintainers' if anyone cares: 1. It's infuriating that The Linux Foundation won't come forward with a clear policy and (better yet) a justification for it. They're putting lead project developers into a bad position of removing people from the project and/or rejecting their contributions without giving them a clear reason, in a way that looks like arbitrary personal decisions, because they've been clearly instructed by LF lawyers to keep their mouth shut. I'm surprised neither Greg KH nor Linus seem to recognize the resulting ill-will. There's more than the linux kernel that is affected by this, and I'll press for that policy if/when there is a right moment, even though I don't hold any formal positions in any of LF projects. 2. That said, I agree with the substance of this action. There must be consequences and cost to military aggression and war crimes. It's equally infuriating that all the defenders and apologists for those russian maintainers won't mention things by name and refuse to get into specifics. 3. I also have little sympathy for those maintainers if I'm perfectly honest. They had all the time they need to find out that their employers are under sanctions, the reasons for those sanctions, what those sanctions mean in practice (given that the kernel is a LF project, and LF is based in California, US), and what are their options. As kernel experts, they are highly sought after on the job market, all the CV they need is that entry in MAINTAINERS file (from past kernel versions, as of now), and it's not that hard to dissociate oneself from the russian war machine. Greg KH made that clear. Instead, they're complaining about being treated poorly, or lack of visibility or other tangents. All of this while Russia continues to fire rockets at civilian settlements and energy infrastructure. Give me a break, pretty please. P.S. I was slightly worried that Linus Torvalds is getting mellow in his middle/old age, from his very diplomatic, careful phrasing at the Linus/Dirk keynote in Vienna. Not to worry, his take on this is 'vintage Linus', brutal and to the point. Check the link below. https://lnkd.in/d5bVgDSE #linux #opensource #linuxfoundation #bombingcivilians

61

8 Comments

Want to come and work with me? We're hiring for a mid-level, a senior and a tech lead software engineer here in Manchester. We are a Scala shop so experience there would be great, but we're also honestly open to smart people with experience in other languages. * Mid-level https://lnkd.in/etE2_err * Senior https://lnkd.in/ehdGirbx * Lead https://lnkd.in/emh_R3Gx Update 2024-04-29: sorry, no visa sponsorship available for these roles. No agencies. Look, I mean it. I know there are arguments both ways but it's not my decision and if you think telling me how good your agency is will change anything, you misunderstand how enterprise companies work.

198

34 Comments

We are happy to announce the launch of an official Rust wrapper for OpenFHE (https://lnkd.in/esNWJxnx), a joint project of Fair Math, Inc and OpenFHE (http://OpenFHE.org) teams. The Rust interface for OpenFHE will significantly expand the capabilities for developing FHE applications, particularly within the #Web3 community, where Rust is a popular programming language. The OpenFHE Rust interface will also be supported by the FHERMA challenges platform (https://lnkd.in/ezVyZ8Nm), which focuses on blockchain and ML applications of FHE. The new Rust interface complements the official C++ (https://lnkd.in/ej2UHuuD) and Python (https://lnkd.in/eXBDsgqE) interfaces, which are already available for application developers interested in applying OpenFHE for their privacy-preserving applications. Thank you Gurgen Arakelov, Nikita Kaskov, Hovsep Papoyan, and Dmitry Suponitskiy for such a nice contribution! Kurt R Rohloff OpenFHE #openfhe #privacy #homomorphicencryption #blockchain #fhe

74

A new blog post: "Proptest: property testing in Rust" https://lnkd.in/dhnYiwBX In the post, I give a real-life example of using property testing in Rust with Proptest with step-by-step guide. #Rust

15

1. Gradle Inc. Gradle is like a magic builder for your software projects. Think of building a house - you need to assemble bricks, wood, and other materials. Gradle does this for software. It takes your code, puts all the pieces together, and turns it into a working app or software. It's super fast and can handle big projects with ease, making sure everything is in the right place. 2. Phabricator Phabricator is like an all-in-one project hub for software development. Think of it as a digital workshop where your team can plan, build, and manage projects together. It helps you track tasks, review code, and collaborate in one place. This makes sure everyone stays on the same page, work moves smoothly, and projects are completed efficiently. 3. Gitea Gitea is like a digital storage locker for your code, but with teamwork in mind. Imagine working on a project with friends. You all need to keep track of who did what and make sure you're not messing up each other's work. Gitea helps you do that by letting you save your code, track changes, and collaborate seamlessly. Plus, it’s lightweight and super easy to set up! 4. Huginn Huginn is like your personal internet assistant. Imagine having a smart robot that watches the internet for you and takes action when something happens. Huginn can send you alerts, gather information, and even perform tasks automatically. Whether it’s monitoring news, tracking emails, or managing social media, Huginn makes your online life easier by doing the repetitive work for you. 5. Tabnine TabNine is like an intelligent assistant for coding. Let's say you’re writing a long essay, and someone whispers the next sentence to you. That’s what TabNine does for coding. It predicts what you’re about to type next and offers suggestions, helping you code faster and with fewer mistakes. It’s like having a smart friend who knows exactly what you need. 6. Ansible by Red Hat Ansible is like a remote control for your IT tasks. Imagine having to cook dinner for 100 people in different kitchens. Ansible helps you control all the stoves from one place. It automates tasks like setting up servers, deploying apps, or managing systems, so you don’t have to do it manually. It’s super efficient, reliable, and saves a ton of time. Follow For More 🤖

6

2 Comments

Soul-searching... Somewhere in a dusty corner, I have a 10-year old low-end Chromebook, that is unsupported for practical purposes even though the hardware still seems to work well (*). Arguably, that machine is tiny by modern standards. Compared to my desktop, it has 1% of the CPU power, even less of the GPU power, 3% of the RAM, and 0.05% of the storage (that is not a typo). At the same time, it is a 64-bit dual-2+GHz, 4GB/16GB, i.e. an unthinkable beast compared to what I had on my desk when I started my career, and desktops back then were more than capable of day-to-day tasks that wouldn't feel out of place today (**). That got me thinking: how can the same hardware feel both so insignificant and so overpowered? And, even further, I've had to ask myself "is this my fault?" and I'm afraid that it might indeed be, to a small extent. I was at Be where we promoted multiprocessor programming, opening the door for today's multicore processors. I was at Openwave where we enabled heavy web technologies on phones where there would have been far better alternatives. I was at Google where I made Android available as Open Source such that the whole industry would be able to deploy and rely on such a heavy system. I was at Yahoo where I enabled hundreds of mobile engineers to collaborate at scale across dozens of apps, freeing up time to build new features. I was at LinkedIn where I improved productivity for about 1000 engineers, again making it possible to keep adding features in that system. It feels like every step of my professional life has been pushing toward the point where machines that are arguably very powerful are considered waste, because I have helped software become increasingly wasteful. Now, I don't think that we can go back infinitely far in the past. Modern approaches to accessibility, internationalization, security, privacy, all add costs that cannot be fully hidden away. But, at the same time, I have a vague feeling that the tech industry is externalizing the costs of client hardware upgrades. I'm hoping to find a path where I can help correct that. I kinda like the tradeoffs in the recent RISC-V products I have seen. (*) Yes, I have Linux on the Chromebook, but the lone distro that supports sound properly on that hardware is too heavy to fit in the 16GB internal flash, I'll need to move to external storage if I want to go that route. (**) The one feature where old hardware quickly falls apart is support for hardware decoding of video formats. There's a balance to be found between the resource savings of continuously improving video formats and the resource costs of handling said formats. Photo: Intel 8080 processors, early examples. That's way too far back for modern applications. I personally think of the i8080 as the first microprocessor meant for general-purpose computing, earlier designs felt more like fixed-function microcontrollers to me. Photo CC-BY 4.0 The Science Museum UK.

48

23 Comments

🔥 Hot tip The shebang line defines which interpreter should be used to execute a script. For example: ``` #!/bin/fish ``` ``` #! /opt/homebrew/bin/fish ``` This is great, but we cannot send a script defined like that to a friend because their fish executable (or whatever else) can be located under a different path. Use this instead 😉 ``` #!/usr/bin/env fish ``` Learned from fish docs: https://lnkd.in/e-aN-vsM #programming 

5

In the era of LLMs, "sexy" might not be the term, but bag-of-words solutions are still kicking! When the position of a word matters, consider using Positional Vectorizer over TFIDF. Check it out: https://lnkd.in/dDRYFrAX

14

1 Comment

This is an excellent blog comparing the streaming ingest of row-level operation comparison among the LakeHouse formats. https://lnkd.in/dsCN66_5

1

There is a new open weights model on the Hugging Face, called MiniCPM V2.6 by OpenBMB. Of course, I tried it locally on my personal MacBook Pro and the HuggingFace space. I used a receipt from two days ago as the input image and asked some questions about the image, including an attempt to extract all the text from the receipt image and format the output in a JSON format. The model wrote the JSON format correctly, but it was not in the video. Perhaps it was because I didn't provide a few-shot examples for the model. Considering the model's size, it's an excellent model to tinker with. I'm not sure yet about running it on a production scale. Primarily because I didn't have enough time to explore the limits of this model. I'm interested in running it natively on the MacBook using the MLX framework instead of using PyTorch with the transformers module. Anyway, it was my first experience with the model. It's a good first impression. Links to the video: - Running locally on my MacBook Pro: https://lnkd.in/g7Du4jyw - Trying on the HuggingFace space demo: https://lnkd.in/gFXs2Rf2 Try it online at https://lnkd.in/ggDHQp7D See the model card at https://lnkd.in/gZ5PR4i4

94

1 Comment

Build Tech Teams from Session Musicians, Not Rock Stars I studied music, not tech, at ACM - Academy of Contemporary Music. We were told over and over again that our role models were to be session musicians, not rock stars. There was a heavy focus on professionalism and substance over style. Being ready on time, having your equipment setup and ready to go, being adaptable to change. 1 minute late, or taking too long to tune your instrument was not good enough in this world of high standards where recording studio time is limited and expensive. The drama of tabloid Rock'n'Roll life style, ego and self belief was drummed out of us and replaced with the importance of discipline, high standards and dedication to the basics. We were taught how musical legends like Jimi Hendrix, Jimmy Page, Glen Campbell and Elton John launched their careers. They were session musicians first - the unsung heroes who transform raw musical ideas into sonic masterpieces. We were shown how session musicians are the invisible infrastructure of music. They aren't chasing the spotlight; they're masters of adaptation, technical precision, and collaborative excellence. Behind great solo artists from Bowie to Madonna were session musicians, recording and touring. Sometimes uncredited for their work. These artists honed their craft by: * Quickly learning diverse musical styles * Supporting other artists' visions * Delivering flawless performances under pressure Consider the Wrecking Crew, who played on over 45,000 recordings. Carol Kaye , the most recorded bass player in history. Consider Hal Blaine who appeared on 6,000 tracks and 40 #1 hits - without most people knowing his name. In tech, we need more of this mindset: professionals who prioritize team success over individual glory. Session musicians teach us that: * Technical mastery matters more than ego * Adaptability trumps fixed skills * Collaborative intelligence drives innovation Great teams aren't built on rock stars, but on versatile, humble professionals who can seamlessly integrate, solve complex problems, and elevate collective performance. Want to transform your engineering culture? Hire session musicians of the tech world. #TechLeadership #TeamBuilding #Innovation

49

6 Comments