Philipp Wagner

Free and open source silicon, the art and craft of making computer chips with (only) free tools, has come a long way -- and has now reached the magic threshold where it can't be ignored any more. In this talk, we'll do a fast-paced journey through what's needed to go from an idea for a silicon chip to actually producing a real physical one using free and open source tools and building blocks. And since talk is cheap, we'll actually build a chip design during the talk in a live demo. Join to learn more about what's possible today, how free and open source is permeating the established industry, and how you can get involved as well.

In the last decades, producing a "real" silicon chip was an effort that required a large-ish team, a fair amount of money, and inside knowledge that only a small, selected group of individuals and companies had. That has changed: free and open source silicon is here, and it democratizes access to chip design for anyone. Free tools in all areas of chip design, from simulation to synthesis and physical design, have attracted renewed attention and are at a point where they are powerful enough to be used for actual chip tapeouts. The last barrier of entry for prospective chip designers fell with the announcement of the SkyWater PDK two years ago: real-world design rules, required to manufacture chips, are now freely available as well.

In this talk, we'll look at the open source silicon ecosystem: which tools do we have available, how are they working together in a toolflow, and can all of that be combined to produce a working chip?

Open source software has irrevocably changed the way software is developed. Open source silicon is set to do the same: inject new life into the traditionally closed chip development ecosystem, combining ideas from the open source software world with the accumulated wisdom of seasoned hardware engineers.

A discussion about the current state of FOSS HDL simulation and synthesis tools, and possible paths moving forward.

Other people: Tristan Gingold Javier D. Garcia-Lasheras Charles Papon

Traditionally, the goal of digital hardware design has been to produce an ASIC. Ideally one which works perfectly after the first tapeout. Tailored towards this goal are our development and testing processes: strictly following a V-model with separated development and verification teams, long design iterations and code which, once it's known to work, is never touched again. For people coming from the software world, this development approach looks arcane. Where are all the sprints, the agile methods, the quick iterations?

With FPGAs being on the rise and available in more and more cloud data centers and possibly bundled with our next Intel processor, we finally get the chance to cheaply make mistakes in digital hardware designs: no more wasted tapeouts, just a new flashing of the FPGA is necessary to fix a bug.

Join me in this talk for a look at development processes and tools. Where can we build bridges between the software and hardware development world, and where do we have fundamentally different needs?

Digital hardware design, or “HDL programming,” is growing in popularity with the widespread availability of FPGA platforms. Yet, the ecosystem around it needs some work. Tooling, the availability of code, licensing and a scattered community make it hard to get started. In this talk, we'll look at the situation today, and where things are going. As newcomer you'll leave the talk with all you need to know to start an FPGA design. And as experienced developer you will get an update on what's going on around you. Because digital hardware design doesn't need to be hard.

Build your own processor! Improve your graphics chip! Sounds far fetched? Actually, it does not need to be.

In the free software world, we have come to enjoy an ecosystem of excellent code to build on, production-quality libraries, reliable tools, and a healthy and helpful community. Software development has become available to the masses.

Getting started in digital hardware design, a.k.a. “writing code that describes how a chip works”, is not as easy yet. Trouble already starts with the question “What do you want to do with your design?” Producing a chip still is rather expensive, and looking at the product of long nights of work in slow-motion simulations it not terribly rewarding. Here FPGAs, “reprogrammable chips,” have come to rescue. With eval boards starting at under $100 and Intel incorporating FPGAs into their processors, hardware access has become easier. With this problem out of the way, we can now focus on the ecosystem of free and open source digital hardware design.

Join me for a look at code, tools, licensing, and the community. What's already there? What can we take from the software world? And where is all this going? All to answer the ultimate question: when will we be able to create our own processor?

This talk introduces OpTiMSoC, a set of open source building blocks to create your own System-on-Chip, which then runs on an FPGA or can be simulated on a PC. The system is formed by tiles like processors or memories connected by a Network-on-Chip, all written in Verilog and supported by a set of software required to run it out of the box. The talk shows how you can use OpTiMSoC to gain insight into a complex System-on-Chip, to evaluate the benefit of new hardware accelerators, or to compare different multicore hardware architectures.

Have you ever wondered what exactly is happening inside a System-on-Chip, like the one that's inside your cellphone, your washing machine, or your car? Have you ever thought "wow, this algorithm could be implemented so much faster in hardware"? Do you do research on embedded hardware architectures and need a platform on which you can try out your ideas?

Then we might have something for you: OpTiMSoC. The "Open Tiled Manycore System-on-Chip" is an open source collection of building blocks to create your own System-on-Chip (SoC). It is based on the concept of "tiles", elements like CPUs, memories, hardware accelerators or external interfaces, which are connected by a Network-on-Chip. On top of those hardware components, which are written mostly in Verilog and can be synthesized to run on an FPGA as well as simulated on a PC, OpTiMSoC contains all necessary software components to get a SoC up and running (a basic "operating system", a C library port, debug support). Of course some running demo systems are included, so you can get started easily by modifying existing designs.

OpTiMSoC was designed by researchers at the Technische Universität München (TUM) and is used there to evaluate new hardware architectures. Over the last two years it grew massiveley in scale and we think it's now useful to others as well. By providing a signal-level insight into a complex SoC it serves as teaching tool and as experimentation and research platform at the same time.

This talk gives an overview over OpTiMSoC, how we use it and how you can make use of it to realize your own ideas.

In the form of XULRunner, the Mozilla Platform can be used as framework to develop powerful applications. This talk looks at how XULRunner together with the Mozilla XForms extension has been used by a university hospital to build a powerful medical documentation system. A special look will be given to the changes over the years and how the Rapid Release Process influenced the way the application and the XForms extension are developed.