Julian Kunkel

Breaking the AI sound barrier with Doctor Syn

At the heart of every new technology, we find it a challenge to get more from less, to optimize. This has proven even more true in the emerging technologies AI and Machine Learning, where the demand for efficiency and optimization is driven not just by a need for high-performance systems, but for squeezing every drop of available computing power out of smaller, more limited devices at the edge.

Despite the demand, many of the core “optimized” libraries upon which AI applications are based were written many years ago, and are not designed to be able to take advantage of modern computer hardware and compiler technology. This has created an AI “sound barrier” that has yet to be broken in which AI code is limited to a certain performance.

In our upcoming talk at the next BCS meeting (16th December), we discuss our new library Doctor Syn, which creates optimizations of arbitrary math functions, by fully advantaging SIMD, multi-threading, and auto-vectorization. We achieve 30x or more speedups over traditional libraries in C, C++, Rust, and FORTRAN, without making the code platform or language-specific.

The AI Sound Barrier

While great effort has been expended in key function optimization, current techniques are unable to efficiently utilize modern compiler technology in order to take advantage of all of the opportunities for parallelism that are available. This leads to the situation in which users are forced to make unfavorable trade-offs between accuracy and speed, with no implementation that is both fast and accurate.

The result is a compromise that has been the status quo for several decades now and, ultimately, there have been very few improvements to math libraries since apart from hand-written assembler versions of older functions. This has created the eponymous AI “sound barrier” of this blog post that has yet to be broken.

Breaking the AI Sound Barrier

The problems that underlie the AI sound barrier run deep. Existing functions struggle to leverage available modern technologies properly because of shared or static libraries, branches, or lookup tables, and, in many cases, fundamental incompatibility. Doctor Syn solves these problems, not by iterating on existing technology, but by leveraging a fundamental rethink of function optimization that utilizes a generic form of optimal polynomial function approximation. This polynomial approximation is in a form that guarantees the technologies of SIMD, multi-threading, and auto-vectorization can be maximally exploited.

Doctor Syn is a technology that is platform and language agnostic, and, both inside the realm of AI and out, can be generically applied to any function with a well-defined domain.

You can hear more about our Doctor Syn by attending the upcoming talk on 16th December, read more about the technology in our blog here, or get in contact with the authors (Embecosm and Atomic Increment) at william.jones@embecosm.com or andy@atomicincrement.com.

Affiliation with Rust London Community

The BCS Open Source Specialist Group (OSSG) is excited to affiliate with the Rust London Community and our new affiliation and support page.

The OSSG is a body of IT professionals consisting of more than 1,500 open source specialists from industry and academia. The Rust programming language is a community project with a vibrant open source ecosystem that brings together the performance of C with the memory safety of Java. 

With the affiliation to the Rust London community, the OSSG will support the community by co-organizing events, announcing events for the Rust community in its calendar, providing letters of endorsement, and giving access to facilities and funding on demand. The first jointly organized event will be revolving around “Embedded Rust” in August.

Statement from the Rust London Community

The Rust London Community is pleased to be affiliated with the BCS Open Source Specialist Group. The Rust programming language has the ethos of empowering through collaboration in its DNA. We here at Rust London are excited to open up our community and connect with the OSSG.

The Rust ecosystem keeps growing and maturing every day. The positive moves forward are reflected in the 2020 Stack Overflow Survey. Rust was voted the most loved programming language for the 5th year in a row, and this is also reflected in the growth of the programming language’s adoption commercially. Several companies and community projects are starting to use Rust more extensively, from web development to space satellites, embedded devices, and game engines. 

Rust London is a place where these companies and individuals can showcase what they are currently doing, what they aspire to do with the language, and how they will positively contribute to the Rust ecosystem. We have grown the community to over 2100 members on Meetup.com, and have more than 700 followers on Twitter. 

The OSSG and the Rust London Community are both excited about what our friendship and connection will bring.

Grant-funded Competition for Development of the ‘Digital Security by Design’ Software Ecosystem

Robin Kennedy, Knowledge Transfer Manager – Cyber Security, KTN

Innovate UK currently have a grant-funding competition open for one more week which may be of particular interest to the open-source community. It’s part of the Digital Security by Design (DSbD) challenge which is investing in projects that help the UK digital computing infrastructure to become more secure.

Development of the ‘Digital Security by Design’ Software Ecosystem

An opportunity for SME stakeholders from across the software development spectrum to explore and investigate requirements, dependencies, and a range of potential complexities associated with the adoption of the Digital Security by Design technologies.

Full competition details are here – the Dates tab includes a link to the recording of the briefing event from late November.

The competition headlines are

  • Only open to applications from UK registered SMEs
  • Fast-Start Short-Term Projects – six months of maximum project length
  • Total Eligible Costs per project : £40,000 – £80,000
  • Closing Date NEXT WEEK – Wednesday 13th January 11:00am
  • Total funding available £1.5 million – will be paid de minimus (i.e. 100% funded)

More background on the Digital Security by Design (DSbD) challenge is here www.dsbd.tech and on Twitter @DSbDTech

COVID19 contact tracing apps: a call for open source

A comment by Julian Kunkel, Simon Worthington, Jeremy Bennett, Andy Bennett

Governments worldwide are developing smartphone apps that track the location and movement profile of citizens in order to quickly identify contact persons of COVID-19 infections. According to The Financial Times, if even 40% of smartphone users install such an application, the infection levels would be significantly reduced in the UK. Therefore, the widespread usage of such an application is an important instrument in the current crisis.

How could such a smartphone app work? In a nutshell, a device can scan other nearby devices and exchange device IDs, for example, using Bluetooth. This information then needs to be stored with a timestamp. If the owner of a device contracts the virus, s/he could indicate this fact in the app allowing to associate the own device ID with the information that s/he may have infected others. This data then needs to be recorded on a server to allow the app of other users to query the register and then compare any contact information with the register of COVID-19 victims.


Case study: The Importance of Open Source for Research in High-Performance Computing

This post is part of the OSSG series “the role of open source in the UK”, where we collect and publish statements from companies and individuals in the UK regarding their experience with Open Source Software. We welcome any submission to this series. If you are interested, please send an email to Dr Julian Kunkel.

by Dr Julian Kunkel, Lecturer, Department of Computer Science, University of Reading

Open source is vital in providing teaching, in conducting research in computer science, and in enabling reproducible large-scale experiments in computational science that support the society. In this post, Julian describes his experience with Open Source in his career.

The Relevance of Open Source: A Personal Statement

Open-source software is for me the key enabler for productive work and for providing training and research environments for various reasons. Firstly, in my own work environment, I rely upon Ubuntu as the operating system to give me the freedom to conduct research and programming experiments easily on my laptop that can later be scaled up to data-center wide experiments. 

Having full control over the system and easy means to repair a broken system, I haven’t lost any data in my 20-year usage of Linux albeit my work often requires to perform rigorous stress-testing of hardware components. I have high confidence and trust in the software stack due to the openness of the software stack. There are no hidden data transmission of private data and proper security schemes in place that protect my data and research. Another benefit I acknowledge is that key APIs are robust and software I rely on that has started to be developed 20 years ago can still be used.


Open Source AI – April 2019

A full-day Open Source AI workshop at April 3rd.

Artificial intelligence will reach human levels by around 2029. Follow that out further to, say, 2045, we will have multiplied the intelligence, the human biological machine intelligence of our civilization a billion-fold.” — Ray Kurzweil

Artificial intelligence promises to aid and augment humans in all facets of our life. As the decisions made by an intelligent system may have wide implications, ethical questions must be resolved as the technology development progresses. Open RD&E can help to increase the trust and reduce the risks.

The slides are now available.

For further information, visit the event page.