Thanks for waiting for 9 years!
Yes, it has been 9 years since we sent the last newsletter.
It was also 9 years since we regularly updated the blog, which was the main source of subjects in the newsletter. Since the beginning of this year, 10 blogs have been posted of which 8 are technical. The latest 2 will be discussed in the next newsletter, else this one will be too full.
In the past year we've also relaunched The Amsterdam HPC Meetup, where we visited the Blender Institute and the Dutch Supercomuting Facilities, SURF. You might have noticed more activity on LinkedIn as well, and seen our articles mentioned on Github, Reddit, X, Mastodon and many more. Good to be back!
The main reason we did not spend enough time on blogging and sharing information: we grew close to 40 employees this year. This means we're still the largest team of GPGPU-developers out there. With a larger team, the projects could get larger as well; as an example we built a part of the compute used by various AI running on the Microsoft Azure Cloud. And all of this was quite demanding. But here we are.
Back to this newsletter. We'll start with a short overview of why and for whom the first 6 technical blog-posts of this year are interesting. Then we'll share about the events where you can meet us. Next newsletter we'll do a bit more, which is discussed in the end. Enjoy!
Six recent technical blog-entries
Your time is limited and therefore an overview to quickly decide what blogs would be interesting to you.
- About: Implementing neural upscaling with convolutional networks in Vulkan compute shaders.
- For: GPU developers and HPC engineers working with Vulkan.
- Why: Learn how to leverage Vulkan for advanced image processing tasks.
- About: Using C++26 static reflection to validate GPU kernel layouts, generate SoA types, and optimize memory bindings.
- For: C++ and GPU developers working on kernel optimization.
- Why: Discover how reflection simplifies GPU code development and reduces errors.
- About: Architectural differences and practical examples of behavior between AMD’s RDNA and CDNA GPU architectures.
- For: GPU developers targeting AMD GPUs.
- Why: Understand the strengths and use cases of each architecture for better optimization.
- About: Exploring
std::execution, C++26’s execution control library for asynchronous and parallel programming.
- For: C++ developers working on concurrent and parallel applications.
- Why: Learn how to use senders/receivers, schedulers, and cancellation for modern C++ concurrency.
- About: Using
std::generator in C++23 to create lazy, on-demand sequences for efficient data processing.
- For: C++ developers working with ranges and coroutines.
- Why: Simplify the creation of infinite or stateful streams without upfront computation.
- About: Using iterators in CUDA and HIP to simplify and optimize parallel primitives like scans, reductions, and sorts.
- For: GPU developers working with CUDA and/or HIP.
- Why: Improve performance and readability by leveraging iterators for complex GPU operations.
Events
You can meet us at various events the coming months. Next newsletter we'll have more events confirmed.
- Data Expo, Utrecht (NL), 9 September.
Without data there is no need for compute. But the other way also is 100% true, and that's why we're there.
- Future Pulse Robotics, Eindhoven (NL), 15+16 September.
Robotics is one of the industries we're focused on. With a fast-growing need for good software for robotics, we're now in the middle of it.
- ESA Industry Space Days, Noordwijk (NL), 17 September.
As Space science and industry is very important for us, we'll be looking for EU-collaborations and of course discuss all the performance engineering opportunities with space companies.
- BSBF, Maastricht (NL). 27-30 October.
Here all the big science of the EU comes together, and we'll be welcoming them at our booth.
If you want to meet us at an event or discuss anything on any another day, schedule a meeting, or email us - you can reply to this email or use the contact form.
Next Newsletter
We'll be discussing a bit more about our plans for 2027 that can be made public, several new "should I read this blog post" overviews, and showing our new homepage. As we experiment with what can be best shared via this medium, what is a good design, how long it should be, etc, there will be some experiments as well.
Got comments? Want to discuss anything? Have a question to one of the authors, but prefer not to ask via the website-comments? We read email you send to us! So just reply to this email with anything on your mind.
See you next month!
(Got this newsletter forwarded? Subscribe here)
