Happy to announce that with the latest changes merged, both messenger types in the MVVM Toolkit are now completely allocation free on .NET 6! They are also both by far the fastest messenger/event-aggregator types across any other MVVM library that I could find! 🚀 #csharp #dotnet

Kevin Gosse Very cool use case! We now publish the ILCompiler NuGet packages to the .NET 7 feed - the dotnet-experimental is now "legacy": github.com/dotnet/runtime…

Got the HelloWorld and DynamicLibrary bflat samples running on my Android phone. New version of the bflat compiler coming soon! Publishing C# libs & command line apps is going to be as easy as 'bflat build --os linux --libc bionic --arch arm64', no dotnet or Android SDKs needed.

I've released a new version of bflat, my C#-to-native compiler driver based on .NET 7. This release extends platform support (ARM64, including Android) and adds more options to make the produced selfcontained programs smaller. Try now: github.com/bflattened/bfl…

Bflat can now create native libraries for Android without requiring the Android SDK. Should in theory be able to build apps that are purely a NativeActivity (no Java). But: is it possible to build an APK without the SDK? I'm guessing I'll need a zipalign and some signing tool?

New version of bflat, my take on a selfcontained C# to native code compiler is out. This release adds an extra base class library that is an extreme subset of .NET. One can't build much with it, but one _can_ build a C# snake game into ~10 kB of selfcontained native code with it

New version of bflat, my take on C# with Go-like tooling is out. New in the 7.0.1 release: build C# apps that run on bare metal hardware without an OS. flattened.net

We can build a completely distroless .NET program using NativeAOT, and even statically link the assembly code into the binary and call it with P/Invoke. Statically linking glibc only yields a 2.8mb size of binary, and it's only 1.6mb without statically linking glibc. #dotnet

Since everyone’s in the game I might as well announce now. I’ve built a WebGPU backed implementation of PyTorch in typescript. Both training and inference using the GPU are supported. It’s available on NPM right now! github.com/praeclarum/web…

Miguel de Icaza ᯅ🍉 I learned touch typing during the pandemic, thanks to Kilo Loco: x.com/peterfriese/st… First, used typing.com to learn the basics Then, improved speed with keybr.com Finally, regular practice with monkeytype.com

WOW! After so many commits and 2 years too late, we finally are using the new things in SkiaSharp: github.com/mono/SkiaSharp… Thanks to Maxwell Katz for making this go green! Waiting for CI to build and then it will appear in the preview feed for your testing pleasure!

If a library uses reflection it doesn't mean it won't work with trimming. Trimming just might need a bit of help. I've submitted a PR with a fix. If accepted, it will be possible to use PDFSharp with trimming/AOT and the apps will be _tiny_! github.com/empira/PDFshar…

If you allocate any native memory on Windows or use lib that does (e.g. anything with C bindings, db, etc) Ensure exe is built with an app.manifest containing <heapType>SegmentHeap</heapType> By default Windows gives you a slow allocator, fastpath it github.com/NethermindEth/…

🚀 OpenResty 1.27.1.1 is out now! Major updates include: • Nginx core upgraded to 1.27.1 • LuaJIT updated to 2.1-20240815 • OpenSSL upgraded to 3.0.15 • PCRE upgraded to 10.42 • New features in lua-nginx-module and stream-lua-nginx-module • Performance optimizations in

You can now train LLMs 3× faster with no accuracy loss, via our new RoPE and MLP kernels. Our Triton kernels plus smart auto packing delivers ~3× faster training & 30% less VRAM vs optimized FA3 setups. Train Qwen3-4B 3x faster on just 3.9GB VRAM. Blog: docs.unsloth.ai/new/3x-faster-…

It is time. Let's go faster! Let's use io_uring for dotnet System​.Net.Sockets on Linux! PR124374 Replaces epoll with managed C# io_uring engine: batching, multishot accept/recv, zero-copy send, registered buffers, SQPOLL (zero-syscall), and more 🚀

Blogged to deep dive into the implementation of Satori GC and compare it to other GCs: dev.to/hez2010/satori…