38C3 - The Design Decisions behind the first Open-Everything FABulous FPGA

  Рет қаралды 4,180

media.ccc.de

media.ccc.de

Күн бұрын

With the availability of robust silicon-proven open-source tools, IPs, and process design kits (PDKs), it is now possible to build complex chips without industry tools. This is exactly what we did to design our first open-everything FABulous FPGA, which is an example of open silicon that is designed and programmed entirely with open tools.
Produced in the Skywater 130nm process node, our chip features 672 LUTs (each with 4 inputs and a flop), 6 DSP blocks (8x8 bit multipliers with 20-bit accumulators), 8 BRAMs (with 1KB each), and 12 register file primitives (each having 32 4-bit words with 1 write and 2 read ports). The resources are sufficient to run, for instance, a small RISC-V system on the fabric. The FPGA comes with a small board that is designed to fit into an audio cassette case and that can be programmed directly via an USB interface. Moreover, the FPGA supports partial reconfiguration, which allows us to swap the logic of parts of the FPGA while continuing operation in the rest of the chip.
The chip was designed with the help of the versatile FABulous framework, which integrates several further open-source projects, including Yosys, nextpnr, the Verilator, OpenRAM, and the OpenLane tool suite. FABulous was used for various embedded FPGAs, including multiple designs manufactured in the TSMC 28nm process node.
The talk will discuss and analyze differences and similarities with industry FPGAs and dive into design decision taken and optimizations applied to deliver good quality of results (with respect to area cost and performance). The talk will highlight state-of-the-art in open-source FPGA chip design and provide a deeper than usual discussion on the design principles of these devices.
The talk will target both FPGA novices and experts and discuss the technology from two angles: 1) the capabilities of open tools to build an entire FPGA ecosystem and 2) FPGA technology insights.
Dirk
events.ccc.de/...
#38c3 #HardwareMaking
Licensed to the public under creativecommons...

Пікірлер: 10
@Jononor
@Jononor Ай бұрын
Super talk. Very dense in information about FPGAs, the key challenges to developing them, and the work on open-source FPGAs.
@Fernando-c2r1w
@Fernando-c2r1w Ай бұрын
Not for beginners, I would say. But quite some great in depth information for me having worked for a few years with FPGAs. Would love to see more such talks at C3. Looking forward to get hands on their chips.
@josvandenanjerklucht
@josvandenanjerklucht Ай бұрын
Gans Geile Talk JA!
@sucim
@sucim Ай бұрын
Great talk!!
@GermanMythbuster
@GermanMythbuster 28 күн бұрын
lave zhe german accent, jaa ❤❤ But for real such a cool talk
@rusus767
@rusus767 Ай бұрын
verry coool talk jaaa
@jean-francoiscaron5706
@jean-francoiscaron5706 Ай бұрын
Fell off the flow at about 20:00.
@zu_alt_fuer2272
@zu_alt_fuer2272 Ай бұрын
ja, me liked, ja
@felixfarquharson
@felixfarquharson Ай бұрын
Certain 2um 1980''s processors *might* be able to cope with a graphical OS like Contiki that have an IPv6 netowking stack, please anyone let me know if you have a link to the source that can be used to make them on home lithography devices like this kzbin.info/www/bejne/iKa5hGqDqLaears
38C3 - IRIS: Non-Destructive Inspection of Silicon
57:55
media.ccc.de
Рет қаралды 2,5 М.
The Genius of RISC-V Microprocessors - Erik Engheim - ACCU 2022
1:01:17
ACCU Conference
Рет қаралды 95 М.
How to treat Acne💉
00:31
ISSEI / いっせい
Рет қаралды 108 МЛН
СИНИЙ ИНЕЙ УЖЕ ВЫШЕЛ!❄️
01:01
DO$HIK
Рет қаралды 3,3 МЛН
WE GOT ACCESS TO GPT-3! [Epic Special Edition]
3:57:17
Machine Learning Street Talk
Рет қаралды 389 М.
Architecture All Access: Modern FPGA Architecture | Intel Technology
20:48
Engineering Intelligent Products and Systems
49:53
Semiconductors @ Purdue
Рет қаралды 58
38C3 - From Simulation to Tenant Takeover
29:56
media.ccc.de
Рет қаралды 19 М.
Coding Adventure: Rendering Text
1:10:54
Sebastian Lague
Рет қаралды 844 М.
GEOMETRIC DEEP LEARNING BLUEPRINT
3:33:23
Machine Learning Street Talk
Рет қаралды 384 М.
A Chronicle of the Unix Wars
22:04
Asianometry
Рет қаралды 227 М.
Deep Learning: A Crash Course (2018) | SIGGRAPH Courses
3:33:03
ACMSIGGRAPH
Рет қаралды 3,3 МЛН
How to treat Acne💉
00:31
ISSEI / いっせい
Рет қаралды 108 МЛН