lowRISC project ideas for GSoC 2015

lowRISC is taking part in the 2015 Google Summer of Code as a mentoring organisation. We are working with a number of our friends in the open source hardware community, acting as an ‘umbrella’ organisation to provide a wider range of projects. For an introduction to GSoC, see the GSoC FAQ.

Students will be able to start applying on the 16th of March. The GSoC Student Guide is a great resource on what is involved in being a GSoC student. Now is the time to look at the suggested projects or consider your own ideas. Please do discuss the project with the proposed mentor. If you have an idea not on this list you want to discuss, please either contact asb@lowrisc.org or post to the lowrisc-dev mailing list.

Key dates:

  • 16th March: student applications open
  • 27th March: student applications close
  • 15th April: Google lets organisations know how many slots they have
  • 27th April: Accepted student proposals announced
  • 25th May: Students begin coding
  • 3rd July: Mid-term evaluations deadline
  • 24th August: firm ‘pencils down’ date

Project ideas (in no particular order)

Your project here: If you have a project idea relevant to lowRISC, don’t worry that it’s not listed here. For a good student with an interesting project we’ll almost definitely have an appropriate mentor. You are strongly recommended to get in touch either on the mailing list or directly to discuss the idea though. Some projects might be better handled under a different mentoring organisation, e.g. a PyPy port to RISC-V would make more sense under the Python Software Foundation.

Schematic Viewer for Netlists (SVG/JavaScript)

Summary: Write a JavaScript library that fills an SVG element with an interactive schematic view for a circuit that is provided by a netlist in a JSON format.

Although usable independent from Yosys, the main application for this will be in educational Web tools for teaching digital circuit design, that are based on YosysJS. There it will replace an ad-hoc display mechanism based on GraphViz. Unlike the existing GraphViz solution, the new schematic viewer should render proper schematic symbols for flip-flops and gates and enable users to modify the placement of symbols (by moving them around interactively). Additional goals would be integration of more advanced features that would directly benefit the educational users, such as the ability to evaluate the circuit with given input values (in this case the JSON netlist would also include evaluable models of all circuit elements, such as an and-inverter-graph).

Links:

Details:

Skill level: intermediate

Language: Javascript

Mentor: Clifford Wolf clifford@clifford.at

A fully open source FPGA compilation flow using Yosys

Summary: Implement a place-and-route mechanism targeting the iCE40 FPGA.

The bitstream protocol of the iCE40 FPGA has been reverse engineered, which means we’re almost at the point where we can have a completely free toolchain targeting an FPGA. The iCE40 provides a relatively small number of LUTs (up to a few thousand), but is very low power. Netlists for the iCE40 FPGAs can be created with Yosys, but there is no place-and-route tool at the moment. This project will involve lots of algorithmic work. A good starting point would be a simulated annealing based P+R. This project has future applications to lowRISC in the case that we include some amount of programmable logic for the ‘minion’ cores.

Links:

Details:

Skill level: advanced

Language: C++

Mentor: Clifford Wolf clifford@clifford.at

Porting Icarus Verilog to JavaScript using Emscripten

Summary: Icarus Verilog is an Open Source Verilog Simulator consisting of two parts: (1) a compiler front-end that creates a vvp-scripts and (2) a runtime that executes those vvp scripts. This project is about porting Icarus Verilog to JavaScript using Emscripten and creating a JavaScript wrapper to easily interface with Icarus Verilog from JavaScript. This project will be used as part of a larger effort to create web-based tools for teaching digital design techniques.

Links:

Details:

Skill level: beginner/intermediate

Language: C/C++/JavaScript

Mentor: Clifford Wolf clifford@clifford.at

Accessing the OpenCores ecosystem

Summary: Implement a Wishbone to TileLink bridge

The OpenCores and OpenRISC community provide many useful IP cores which have been used in ASICs and FPGAs for more than ten years. Most of these cores use the Wishbone interface for communicating with a CPU. This project would create a tested and documented bridge component between TileLink and Wishbone to allow these cores to be easily integrated in the lowRISC project and by others in the RISC-V ecosystem.

Links:

Details:

Skill level: intermediate

Language: Verilog

Mentor: Olof Kindgren olof.kindgren@gmail.com

Adding Chisel support to FuseSoC

Summary: Extend FuseSoC to support Chisel

FuseSoC is a package manager and build system written in Python with the intention of making it easier to reuse cores from different sources, combine them into a SoC and run through simulators and FPGA/ASIC flows. Currently, only Verilog and VHDL (and C/C++ for simulations) is supported. This project would also add support for cores written in Chisel such as the Rocket core, which would eventually allow FuseSoC to be used as the build system for the lowRISC SoC.

Links:

Details:

Skill level: beginner/intermediate

Language: Python

Mentor: Olof Kindgren olof.kindgren@gmail.com

Extend Tavor to support directed generation of assembly test cases

Summary: Tavor is a framework for generation-based fuzzing which can be utilized to quickly provide an initial set of tests for an implementation. In comparison to formal methods, such as model checking, it is not meant to perform a full verification. Instead, fuzzing can be applied to test if an implementation is working at all.

The goal of this project is to use and extend Tavor to fuzz instruction sets. It should be possible to easily maintain a definition of an instruction set to generate a small amount of assembly tests. Given an appropriate heuristic, these tests should cover all defined instructions in multiple corner cases acting as a sanity test for the implementation under test.

Links:

Details:

Skill level: intermediate

Language: Go

Mentor: Markus Zimmermann markus.zimmermann@nethead.at

Constrained randomised testing with coverage tracking in Cocotb

Summary: Extend the Cocotb co-simulation library to support constrained randomised testing (e.g. via Google or-tools) and to track achieved coverage of the tested HDL.

Cocotb is a Python framework for testing VHDL and [System]Verilog hardware designs. Although various open source simulators are available, none of them provide the advanced verification features of expensive proprietary simulators. This project will implement constrained randomisation and functional coverage collection in Cocotb, giving open source projects these capabilities.

Constrained Randomisation is a similar concept to fuzz testing; generating random stimulus transactions according to certain constraints in order to exercise a hardware design. To understand how well the randomly generated inputs are testing design the we also need to instrument the code to track metrics on which scenarios have been exercised (known as functional coverage).

The goal of this project is to provide a convenient interface to an existing constraint solver from Cocotb and create and manage a database of functional coverage points. To facilitate processing of the coverage data we’ll need to export to various formats for consumption by other tools. We can also integrate coverage information with existing software development services such as coveralls.io.

By undertaking this project you will learn about latest ASIC/FPGA verification practices and the interaction between hardware and software development

Links:

Details:

Skill level: intermediate

Language: Python

Mentor: Chris Higgs chris.higgs@potentialventures.com

LLVM pass for control-flow hijacking protection using lowRISC’s tagged memory

Summary: Implement the control-flow hijacking protection scheme outlined in the lowRISC memo

lowRISC supports tagged memory, a mechanism that associates metadata (tags) with every location in physical memory. The initial motivation was to protect against control-flow hijacking, but it has a number of other potential use cases (described in more detail in our memo). A simple application of tagged memory is to mark every code pointer (such as return addresses or vtable addresses) with a tag making them read-only. The program should also be modified so the presence of this tag is checked upon loading the code pointer. This project would implement an LLVM pass applying this policy. Possible extensions to the project would be analysing the overhead of this policy on various programs, or looking at the protection of heap metadata.

Links:

Details:

Skill level: intermediate/advanced

Language: C++

Mentor: Alex Bradbury asb@lowrisc.org

TCP offload to minion cores using rump kernels

Summary: Use the NetBSD ‘rump’ kernel technology to run an IP stack on lowRISC’s small RISC-V ‘minion’ cores for the purpose of TCP offload.

Rump kernels allows the reuse of NetBSD code such as driver implementations or the TCP/IP stack in other environments. These high quality drivers can be made to run on a diverse range of platforms including Linux userspace, the Xen hypervisor, and bare metal. The goal of this project is to get rump kernels running in a bare metal RISC-V environment. A useful demonstration and proof of concept would be to show the NetBSD TCP/IP stack running on something approximating a lowRISC minion core (e.g. a RISC-V core with a 16KiB instruction cache and a fairly high cache miss penalty). The sophistication of the TCP/IP offload offered will depend on the experience of the student and the time available towards the end of the project.

Links:

Details:

Skill level: intermediate/advanced

Language: C

Mentor: Justin Cormack justin@specialbusservice.com

Porting L4/FIASCO.OC to RISC-V

Summary: Port the L4 FIASCO.OC microkernel and L4 runtime environment to RISC-V.

L4 microkernels are having their revival nowadays as secure and thin layer between the applications and the hardware. The traditional approach of a microkernel is that all kernel services are also executed in the user space, each in one container. Contrary to monolithic kernels, a kernel service is therefore accessed with inter-process messages and executed in user space. Only when necessary, certain capabilities are granted to a service container, e.g., access to a device for a driver service. This allows for strong separation, but with the drawback that the inter-process communication needs to be really efficient. While the original microkernel approaches therefore lacked of success, modern computer architectures are better suited for the demands on the inter-process communication. The L4 family of microkernels is the most popular.

FIASCO.OC is an L4 microkernel developed at TU Dresden/Germany. It is written in C++ and has real-time capabilities, multi-processor support and an object-oriented capabilities system. Similar to hypervisors, it is also possible to encapsulate entire operating systems in para-virtualized containers (such as L4Linux, L4OpenBSD or L4Android). The goal of the project is to port the FIASCO microkernel to the RISC-V architecture. Beside the basic kernel functionality (memory mapping, processor sharing) a set of service containers from the L4 Runtime Environment (L4Re) also need to be adapted to the RISC-V architecture.

Links:

Details:

Skill level: advanced

Language: C/C++

Mentor: Stefan Wallentowitz stefan@wallentowitz.de

Trace Debugging Infrastructure for lowRISC

Summary: Design and implementation of a trace debugging infrastructure for the cores and uncore logic in the lowRISC SoC.

Debugging embedded systems has become more and more complex. Traditional run-control debugging needs to be adapted to be aware of multiple processor cores and other active elements. For this, cross-triggering techniques and sophisticated tools are helpful for example. Beside this, trace debugging techniques become more and more important. Here, the hardware is monitored and certain events are generated, often with the timestamp of the event. For example, each change of the program counter can be trace event or alternatively only branches are considered trace events. Beside this, other hardware or software events may be part of an execution trace. The trace is generated while the system executes with minimal interference and analyzed offline (or also online). The goal of this project is to analyze existing trace debugging approaches, especially the Nexus standard. Based on the discussion hardware monitors are then developed and the configuration and actual tracing infrastructure are prototypically developed for the lowRISC system-on-chip. Especially the different kinds of cores and other elements of the SoC should be considered.

Links:

Details:

Skill level: intermediate

Language: (System)Verilog or Chisel

Mentor: Stefan Wallentowitz stefan@wallentowitz.de

Optimized ray tracer for Nyuzi parallel processor

Summary: Write a ray tracing library that is optimized for Nyuzi parallel processor.

Nyuzi is an open source parallel processor architecture. This project would implement a ray tracer that takes advantage of both vector arithmetic and hardware multithreading. This would act as a benchmark for exploring the performance of this architecture and also a validation test. Part of the project may also involve proposing/implementing instruction set or architecture extensions to improve performance.

Links:

Details:

Skill level: intermediate

Language: C/C++

Mentor: Jeff Bush jeffbush001@gmail.com

JTAG hardware debugging support for Nyuzi

Summary: Add the ability to single step, inspect memory, set breakpoints over JTAG.

Nyuzi is an open source parallel processor architecture. This project would implement support in the hardware pipeline (running on FPGA) for control by a host over JTAG.

Links:

Details:

Skill level: advanced

Language: SystemVerilog, C

Mentor: Jeff Bush jeffbush001@gmail.com

Porting musl libc to RISC-V

Summary: Create a port of the musl libc to RISC-V 32-bit and 64-bit.

Musl is an MIT-licensed libc implementation with excellent support for static linking. It has been used by a number of lightweight Linux distributions, most prominently Alpine. This project would involve porting musl to the 32-bit and 64-bit RISC-V instruction set architecture. An unoptimised port should require less than 1KLOC, so after achieving this initial milestone time should be spent both building out the musl benchmark suite and implementing optimised implementations for key functions.

Links:

Details:

Skill level: intermediate

Language: C

Mentor: Rich Felker dalias@libc.org

jor1k port to RISC-V

Summary: Write a RISC-V 32-Bit CPU emulator in Javascript

jor1k is an emulator for the OpenRISC platform and is the fastest emulator which runs in the web browser and boots Linux. It comes with a working underlying framework and a comprehensive library for hardware devices such as a framebuffer and network support, which can be used for the RISC-V too. This project would involve the programming of the missing part - a RISC-V 32-Bit CPU emulator following the ISA specification (estimated 1-1.5kLOC). A milestone would be the booting of Linux. We will then focus on optimizations using asm.js. Possible extensions include adding support for other features of the lowRISC platform, such as tagged memory or minion cores. Basic knowledge about CPU design and general assembler programming is required.

Links:

Details:

Skill level: intermediate/advanced

Language: Javascript

Mentor: Sebastian Macke sebastian@macke.de

OCaml native code port to RISC-V

Summary: Write a RISC-V 32-Bit native code generation backend for the OCaml compiler

OCaml is a functional programming language from the ML family, with code that can be compiled natively to x86, ARM, Sparc, PowerPC and MIPS binaries using the ocamlopt compiler included with the distribution. This project would add support for the RISC-V instruction set into ocamlopt, and verify that the resulting code generation passes the test suite included with the OCaml compiler. Once that passes, a stretch goal would be to compile libraries in the OCaml ecosystem to test their correct operation using the OPAM package manager.

Links:

Details:

Skill level: intermediate/advanced

Language: OCaml

Mentor: Anil Madhavapeddy anil@recoil.org