std::vector always heap allocates. std::array can't change size. For decades, there's been no standard container that gives you a dynamically sized array with a compile-time capacity limit and zero heap allocation
C++26 finally adds std::inplace_vector. Guess where they got the idea 🧵👇
@MatRopert@blelbach@joaquintdig That double standard isn’t coming from me. Asio is arguably existing practice and, therefore, a candidate for ISO. But I’m not convinced that the benefit is worth the cost. There are benefits to leaving it in Boost.
@MatRopert@blelbach@joaquintdig The mandate of the committee is to standardize existing practice. For language features, this is problematic, but for libraries—nothing should go in that is not already in Boost (or a comparable source). The committee should fine-tuning not inventing.
@MatRopert@blelbach@joaquintdig I agree. We need to stop putting containers (or anything else that ages poorly) in the standard. But we also need to teach people the value of Boost and other premium resources.
@MatRopert@blelbach The standard has laid out the conventions that are the STL. It has a few (too many) containers, iterators, and algorithms. Now go forth and create your own. And if/when they are good enough, put them in Boost. The standard is the wrong place for every library that is a good idea.
I tried to stop std::hive.
When it became clear I couldn't, I tried to keep them for baking in a design that forced a particular implementation.
std::hive doesn't belong in the C++ standard library.
std::queue has no concept of concurrency. Wrap it in a mutex and every push serializes behind every pop: 16 cores fighting over one lock, 15 of them idle at any given time 🧵👇
@dontquillthepod Steven, your strength is an inspiration to us all: “Instead of weeping when a tragedy occurs in a songbird's life, it sings away its grief. I believe we could well follow the pattern of our feathered friends.”
Penn State + Lockheed Martin built a reinforcement-learning cognitive engine for @NASA and the @Space_Station in C++11
The engine manages real-time radio resource allocation across 6 competing objectives: bit error rate, throughput, bandwidth, spectral efficiency, transmit power efficiency, and DC power consumption
Two Boost libraries did the heavy lifting:
Boost.Asio: concurrent UDP listeners and raw Ethernet frame transmission to DVB-S2 modems, within a 40ms ground-to-orbit RTT constraint
Boost.Serialization: complete engine state (NN weights + training buffer) archived to disk between passes, so the radio resumes learning instead of starting over
20 live ISS flight passes. One of the first published cognitive engine experiments on a space-based asset
Learn more https://t.co/TX8W49Cy1C
The ISS passes over a ground station for a few minutes at a time. Your cognitive radio engine has that window to transmit and to learn. How do you preserve neural network state across orbital passes in C++? 🧵👇
Don’t hesitate. The conference itself won’t sell out, but our room block will and being in the main hotel is so much nicer than being in an overflow hotel.
Registration for CppCon 2026 is now open which will be held at the Gaylord Rockies in Aurora, Colorado from 12th-18th September.
https://t.co/ltm5Bbpw6N
Check out our promotional video to find out what to expect!
https://t.co/8ScElaRNMU
#cpp#programming#coding
📣 The Boost formal review of Multi begins today
Multi, by Alfredo Correa, is a modern C++ library for manipulating and accessing data in multidimensional arrays, on both CPU and GPU memory
The review runs March 5–15, 2026. If you’re interested in a first look:
🔗 Code: https://t.co/Z6M9tz1htR
📖 Docs: https://t.co/kk0URouawI
📬 To participate: https://t.co/vC3AZdebuR
#cpp
Boost.Math now supports reverse mode automatic differentiation. Exact derivatives via computational graphs, no step size tuning, no approximation error. Thanks to Maksym Zhelyeznyakov (physics BS, applied math MS, ECE doctorate from UW), who hit this wall on a personal project
Now the same technique that powers gradient computation in PyTorch and TensorFlow is native in C++ through Boost. If you're building ML inference engines, trading systems, or physics simulations in C++, this changes your workflow
#cpp