Welcome to Libre-SOC

We're building a chip. A fast chip. A safe chip. A trusted chip.

A chip with lots of peripherals. And it's a VPU. And it's a 3D GPU...

Oh and here, have the Libre-SOC source code...


Why should I care?

With Libre-SOC, you can take complex algorithms usually intended for power hungry servers with big fat GPUs, and run them on tiny devices like smartwatches, cellphones, and pocket drones without changing your code at all.

Hasn't Somebody Already Done This?

To the best of our knowledge, no. The closest systems would be ARM Cortex devices which currently offer mediocre GPU and OpenCL support. Often times, it is quite difficult for customers to get their hands on the drivers and install them due to their locked down nature. Libre-SOC is providing our own Free/Libre drivers. Easy as 1, 2, 3!

Does Open Hardware Really Work?

A few names come to mind:

Learn more

# Our Team We are proud of our talented and diverse [team](about_us) located all across the world: UK US Canada Germany Greece India Kuwait. # Our sponsors and partners * [NLnet](https://nlnet.nl) who have very kindly sponsored us with over EUR 550,000 in various separate grants. * [NGI POINTER](https://www.ngi.eu/ngi-projects/ngi-pointer/) who have sponsored us with a EUR 200,000 Horizon 2020 Grant. * [Mythic Beasts](http://mythic-beasts.com) who don't overcommit on their VPS allocations (no memory ballooning), and put servers on the fastest backbone Internet points. * [Purism](http://puri.sm) donates to us through [NLNet](nlnet) charitable gifting * [Raptor CS](http://raptorcs.com) has given us access to a powerful 18-core 128 GB RAM TALOS II workstation, online. * [Raptor Engineering](https://raptorengineering.com) is providing additional assistance including access to an [[shakti/m_class/LPC]] interface (more to come) * [MarketNext](http://marketnext.org) is helping us connect to developer resources in Emerging markets, for completion of NLNet-funded tasks. * The [[PowerPC Notebook]] Project * [RED Semiconductor](http://redsemiconductor.com) is the commercial realisation of Libre-SOC technology that respects and actively supports Libre principles in its business objectives. * [[fed4fire]] for supercomputer access for verilator simulations, with thanks to Ghent University / IMEC. # How Can I Help? If you would like to fund us, see [[funding]]. We currently have some funding and always appreciate more! If you are a Corporation or an individual you can donate tax-free to NLNet, and they can in turn gift our developers. Contact lkcl@lkcl.net for more information. # How Can I Help as a Developer? If you want to write code with us (and receive donations from NLNet for doing so), keep reading. If you want to *learn* so that you can write code with us, see "How can I learn" section, below. If there is anything else, just get in touch on the list, there is plenty to do. 1. First, join the [mailing list](http://lists.libre-soc.org/mailman/listinfo/libre-soc-dev), introduce yourself (people will happily say "hello" back) and the Libera IRC channel (#libre-soc at irc.libera.chat port 6697, see https://libera.chat/guides/connect), it's also accessible via Libera.Chat's Matrix Bridge. [logs here](https://libre-soc.org/irclog/latest.log.html). Read through [recent posts](http://lists.libre-soc.org/pipermail/libre-soc-dev/) and the [[charter]], ask questions if you have any and let everyone know, on the list that you're happy with it and agree to it. 2. The next thing you should do is read through the [bugs list](http://bugs.libre-soc.org) and see if there are any bugs that pique your interest. A fascinating way to do that is to view the [dependency graph](https://bugs.libre-soc.org/showdependencygraph.cgi?id=1&display=web&rankdir=LR) 3. After that, go ahead and take a look at the [git repositories](https://git.libre-soc.org). 4. If you plan to do HDL work, you should familiarize yourself with our [[HDL_workflow]]. If you would like to help with the ASIC layout, see [[HDL_workflow/coriolis2]] 5. We do have funding available (see [[nlnet]]) upon completion of issues - we are also working on procuring more funding which gets the project to nanometre scale tapeout. 6. After all this, if you feel that Libre-SOC is a something that you would like to contribute to, add yourself to the [current_members](about_us) page, fill in some information about yourself, and join the mailing list and say hello. Also note that you can edit this wiki. You can experiment in the [[Sandbox]]. ## Quick peek at the code The full install process may be automated using scripts found here: which you can clone with: git clone https://git.libre-soc.org/git/dev-env-setup.git A convenient page which gives easy step-by-step instructions is: [[HDL_workflow/devscripts]], and you can get started with a [[docs/firststeps]] tutorial. ## How can I learn? The whole purpose of this project is to be a learning environment as well as an actual business. If you want to learn how to do ASIC design, with a view to contributing or starting your own ASIC business, start here: [[3d_gpu/tutorial]]. Yes, really: do not let anyone tell you that you can't learn this stuff too. Along the way, if you find any [[resources]] or links that you found particularly helpful, please add them to that page so that other people can also benefit (and you have a place to remember them, too). ## Needed Skills Most labor is currently being applied to developing the GPU portion of the Libre-SOC. Our highest priority short-term and medium-term: we need HDL engineers. Particularly those familiar with nMigen or just python. Most of the techniques being used require software engineering skills (OO design, polymorphism) than they do more traditional HDL programming skills. Basically if you have experience in 2 of the following you'll do fine: python, nmigen, verilog/VHDL/gate-level design. See [[HDL_workflow]] We also need to find a c++ engineer with compiler experience to work on a MESA 3D driver. This will begin life similar to SwiftShader however retaining the vectorisation and predication intrinsics then adding hardware accelerated opcodes (all of which SwiftShader lacks) Also, individuals with experience in formal mathematical verification are quite welcome. # Documentation - [Source Code](/docs/) - [Architecture](3d_gpu/architecture)