Democratizing the use of Heterogeneous Computing to enable Intelligent Products of the Future
Now that the dust has settled from our Series B funding announcement, there is already another exciting event to celebrate today: Silexica’s 4th anniversary! I’d like to provide some insights on where we are as company, where the industry is heading and our focus on creating a new software development platform to bring prototypes enabled by heterogeneous computing to production.
Every startup is a wild mix of fun, enthusiasm, wins and losses, a constant battle with risks and for sure we have made a number of mistakes. But we iterate, learn quickly and adapt and I’m thrilled about the team we have built, the progress we have made and our footprint within various industries. I am so grateful for the dedication and hard work from everyone in our amazing team. They will continue to push boundaries and deliver industry firsts to ensure compute- and data-intensive applications run efficiently on embedded computers.
At the same time, I’m very thankful to our customers for believing in us and I can’t wait to see what else we can achieve together!
Spin-off to today, addressing the shift from single to multicore
We started Silexica in 2014 after forecasting major challenges due to the transition from single processor cores to homo- and heterogeneous multicore platforms.
Some of the key challenges during this ongoing transition are:
1. The selection of an appropriate compute platform
2. The migration (or many times a very expensive rewrite) from sequential to parallel code
3. The distribution of parallel code to a chosen multicore platform based on different optimization criteria and tight timing and power constraints.
Despite hardware innovation coming at an incredible speed, programming tasks are still executed manually using educated guesses, endless trial-and-error iterations or using tools such as Excel. See more thoughts on this from Andrew, our VP Sales, here.
Instead of innovative tooling to help developers utilize these advancing hardware platforms, new programming languages and methodologies have been introduced. This creates more confusion for software professionals rather than helping to solve the problem.
Pictured below you can see an overview of the landscape with various frameworks and programming languages mapped to different hardware technologies. This is not an exhaustive list, but serves as a guide to the current situation.
Our role at Silexica is to support software professionals in unravelling this in compute-intensive industries. At the moment we focus on automotive, wireless and aerospace projects who all operate within their own compute environments and workflows. There is no one-size-fits-all solution and we built a powerful selection of SLX products for various inputs and developer environments. Those are production-proven, extensible frameworks to integrate into the workflow and specific use case. However the challenge is getting bigger…
Multicore is only the beginning
The next wave of hardware innovation pushed by new product requirements is already upon us. Multicore-based products are getting even more complicated and the next generation of embedded computing platforms will see multiple applications with different models of computations running on heterogeneous compute islands as shown in this diagram.
Having witnessed the explosion of complexity during the shift from single core to multicore, this next step is even bigger. With products demanding high performance, safety, resilience, predictability, updatability and faster times to market – new disruptive solutions need to be provided to mitigate putting complete industries at risk.
Let’s look at an example of a central compute platform for a level 3 autonomous vehicle. In fact, this could be any intelligent edge device, where high-compute algorithms, neural networks and safety critical code all run and interact together on the same platform.
As mentioned before, the input source code, developer environment and optimization criteria vary massively among those different clusters. Therefore, we have designed different products integrated into existing workflows, an overview can be seen below. For more details on these products, please visit our Products page.
As we move towards complete driverless cars, higher levels of autonomy (level 4-5) will see even more powerful computers with several platforms connected to each other in so-called “system-of-systems”. Additionally, they will need to link to smart devices outside the car such as other vehicles, road infrastructure or the cloud in a computing everywhere environment.
In such systems, understanding processing needs is simply not enough. This incredible flow of data is extremely complicated to understand, debug and validate – emphasizing the urgent need for system level tooling to give further insights. In addition, the heterogeneity of domains require a much tighter interdisciplinary collaboration of different in-house development teams, suppliers and contractors.
That is where we see an additional opportunity with our tooling and domain experience. We want to build a new software development platform taking intelligent products from concept to mainstream, supporting the design cycle for all the software professionals involved. While I will talk about the various roles during a complete product design cycle in a follow-up blog post, I’d like to give a first glimpse into the SLX.platform.
The Silexica vision: SLX.platform
Looking at all the described challenges, disruptive tooling is the only solution. This needs to be not only a continuation of our existing products to optimize individual applications, but a new system analysis tool usable in a vendor-neutral, collaborative way.
Our new SLX.platform will be exactly this. A next-generation, collaborative development platform and marketplace for the design and optimization of intelligent products of the future. It will be connecting software developers, architects, multicore engineers and hardware providers while protecting everyone’s individual intellectual properties.
As a first step, we are building a new simulation framework which allows unprecedented system insights into the behaviour of multiple applications running simultaneously on complex, heterogeneous systems.
Due to the complexity of future systems, isolated decision making (e.g. software distribution, memory allocation and scheduling) for each individual application will not be sufficient anymore. Therefore, a highly extensible framework will take the output of our individual analysis and optimization tooling and combine it into one unified view of the system to properly capture the interaction among multiple applications.
New insights include the utilization of available computation and communication resources, available headroom and overall system stability. This will allow powerful what-if analysis based on an upcoming open and royalty-free platform description standard. It is important to note that this is meant to run on a continuous basis and not just at the final integration stages of a product cycle.
In order to use those insights in a collaborative manner, while still protecting source code, we are envisioning SLX.platform to combine both cloud and desktop solutions leveraging their respective strengths and advantages. All model-based development can be used directly in the cloud to analyze, optimize and collaborate across team and company borders. And we will utilize our 15 years of deep source code and binary understanding to connect our current desktop solutions to this platform, generating representative abstraction models automatically. A transparent integration between the two worlds will guarantee a natural and seamless use and allow cross-team and cross-vendor system insights and collaboration.
We are only at the beginning and there is much more analysis and implementation work to be done in order to build and ship some of the described solutions. However we are already making great progress.
Please reach out to us to support your use case from the beginning – we want to disrupt the way products are designed in the future together with our customers and partners!
I’m very excited that our Series B lead investor EQT Ventures, one of the leading venture capital funds in Europe, shares our view of the industry and our vision how to support them with new solutions. You can read more about EQT’s investment thesis “the time for Europe is now” here and why they recently invested into Silexica in a short video clip here. The investment round was also discussed in various news articles, including here and here.
All existing investors have joined this round again with significant financial commitments and I’d like to whole-heartedly thank them for continuing to believe in the evolvement of our vision.
We understand very well that raising money is neither the ultimate goal of a company nor any proof of a potential success of a company. But it is a necessary fuel to accelerate the creation of a market-winning company from scratch and build out a value-delivering product line.
The future is now
We will continue to push boundaries and working tirelessly to deliver our promises with the ultimate goal of democratizing the use of next-generation, heterogeneous computing, enabling intelligent products of the future to enrich our lives.
I sincerely hope this gave valuable insights into the multicore challenges of making intelligent devices a reality and how Silexica can help. I’d love to hear your feedback and gain more insights into your own challenges and use cases, so please contact me directly.
Maximilian is on a mission to democratize the use of next-generation computing, enabling intelligent electronic products of the future. He is the CEO and co-founder of Silexica and has built the company from its beginning in 2014 to become a global leader in software design automation for heterogeneous computing. Silexica now has a team of 50 compiler, software, hardware and AI experts based in offices in Germany, Japan and the USA.
Max received a Computer Engineering diploma from RWTH Aachen University in 2010 and was formerly the Chief Engineer of the Chair for Software for Systems on Silicon leading 15 research assistants. He has had over 20 publications published in international computing conferences and journals.