Google Summer of Code 2018 - Ideas for Projects

Whether you’re an aspiring student or mentor, feel free to contact us, either through the private GSoC-specific mailing list gsoc@fossi-foundation.org, through the public discussion mailing list, or through the mentors listed for each project below. We are also available on Gitter in librecores/Lobby.

Looking forward to meet you all!

Create your own LibreCores, or contribute to an existing one

Details: Our main goal is to grow the community around open source silicon designs. LibreCores are IP cores, but they are free and open. While there are many projects you can contribute too, you may have your own great idea for a LibreCore. All projects start small, and we see this is a great chance to bring forward new ideas and start building new tiny bits and pieces that enable free and open source chips.

We are happy to mentor you with your own idea, but it is important that it is re-usable and contains everything needed for simple and flexible integration, like testbenches, the required software drivers etc. So, it is important that you discuss a proposal intensively.

Skill level: All

Language/Tool: Verilog, VHDL, Chisel, …

Mentor: We will find the mentor with you, LibreCores GSoC team

Extend LibreCores.org

LibreCores is a community web site with the goal of providing an overview of IP cores and the corresponding ecosystem. We strive for LibreCores to be the resource for free and open source IP: it should be easy to find, integrate, and contribute to the projects found there – to make digital design projects as easy as writing software. For further information on our goals, see the FOSDEM Presentation slides announcing LibreCores. The full site source code is available on GitHub.

You can find a non-exhaustive list of available tasks in our documentation. Please talk to Philipp if you have other ideas, or didn’t find an interesting project. We welcome your own ideas!

Skill Level: Intermediate

Language/Tools: PHP7 with the Symfony Framework, MySQL, HTML/JS

Mentor: Philipp Wagner

Continuous Integration for Hardware Projects on LibreCores CI

Goal: Setup verification and continuous integration flow for one of open-source digital hardware projects.

Details: LibreCores CI is a under-development Continuous Integration service within LibreCores. In this project we offer students to work with modern hardware verification tools, RTL codebase and Jenkins Pipeline in order to setup efficient verification flows for one of the open-source hardware project being hosted on LibreCores. The project includes improvements of the HW project testability in RTL, development/improvement of testing frameworks and a development of a new Pipeline Library for automation in Jenkins.

Prerequisites:

  • Basic knowledge of the hardware verification techniques
  • Knowledge of one of RTL languages
  • Knowledge of one of the scripting languages (preferably Python or Groovy)

Skill Level: Intermediate

Language/Tools: Verilog/VHDL/…/Python, Jenkins, Groovy

Mentors: Oleg Nenashev, Stefan Wallentowitz

Open SoC Debug: Efficient Control-Flow Traces

Trace debugging is the method to observe the execution of a system-on-chip. The Open SoC Debug project creates open source building blocks for a debug infrastructure, with a strong focus on efficient trace debugging.

Control Flow tracing (or instruction tracing) is the process of observing a program execution on a processor, and creating a “trace” from these observations, which allows a tool to re-create the program flow as it happened on-chip.

To transmit traces efficiently from the chip to the host PC, trace compression must be applied. This compression reduces all non-essential parts of the trace, reducing it to (typically) only jump instructions where the target cannot be determined using the program binary.

The goal of this project is to (a) find a good trace compression algorithm for instruction traces, and (b) implement it in System Verilog (on the hardware side) and C (on the software side) for the Open SoC Debug system.

Skill Level: Intermediate

Language/Tools: SystemVerilog, C

Mentor: Philipp Wagner

Open SoC Debug: Trace Visualization and Configuration

Details: The Open SoC Debug project creates open source building blocks for a debug infrastructure, with a strong focus on efficient trace debugging. Currently we focus on the target (hardware) side of the infrastructure, but want to improve the host software, especially visualization of traces and configuration of the debug hardware.

In this project you should not reinvent the wheel, but build around existing infrastructure. For example the Open Trace Format 2 and the SCORE-P infrastructure are good starting points. For the visualization and interface building we suggest having a look at state of the visualization tools like ravel to integrate with or build a new framework for example on electron.

Skill level: Beginner, Intermediate

Language/Tools: C++/Java/JS

Mentor: Philipp Wagner, Stefan Wallentowitz

Open SoC Debug: Trace Logging to Memory

Details: In the Open SoC Debug we currently transfer traces from the debug target to the host for on-line visualization or offline processing. But low level traces may be interesting even while the system-on-chip is in the field, similar to system traces, e.g. from Linux. The idea is to write the traces to a reserved space in the system memory and read them from the running software.

Basically this idea involves two hardware tasks: A configuration interface for trace logging and the interface between the debug interconnect and the system memory. Ideally your proof-of-concept includes a simple software. This setup can be optimized for example with trace compression and circular buffering.

Skill level: Intermediate

Language/Tools: System Verilog

Mentor: Stefan Wallentowitz

Open SoC Debug: Run-Control Debugging

Details: Open SoC Debug is a free and open debug and trace system for embedded systems. Our current developments mainly focused on the tracing part, leaving run-control debug for this GSoC project! By run-control debug we mean the process of firing up GDB, setting breakpoints in a program, and stepping through the program. In order to support these operations from GDB, the goal of this project is to create (a) hardware extensions in SystemVerilog, which allow reading control registers from a CPU (these are used to set breakpoints and halt/continue the CPU), and (b) create a gdbserver implementation inside our Open SoC Debug software implementation which GDB can connect to. To get you started, we have sample code available which can be used as a starting point.

Skill level: Intermediate

Language/Tools: System Verilog, C

Mentor: Philipp Wagner

OpTiMSoC: Extend the Linux Port

Details: The Open Tiled Manycore System-on-Chip is a prototyping platform for massively parallel multicore system-on-chip. The main runtime environments we employ so far are baremetal and a very lean operating system (“compute node OS”).

In last year’s GSoC Pedro created an initial Linux port for OpTiMSoC. He solved all the hard problems of getting Linux up and running, and paved the path to continue with more high-level features of Linux. By now the OpenRISC port of Linux is upstreamed and extended to provide multi-core support (SMP).

The goal of this GSoC project is to extend the OpTiMSoC Linux port. What extensions, you might ask? It’s up to you! How about enabling multi-core support? Or creating an extended device driver for our Network-on-Chip (NoC)? Or writing an accelerator interface to trigger computations in the compute grid of OpTiMSoC from software running on Linux? There’s an endless stream of opportunities to choose from – bring your own idea or get in touch with us to discuss ideas we have.

No matter which task you choose, you get unique insights into hardware (you can actually view the waveform of most signals in OpTiMSoC!) and Linux internals.

Skill level: Intermediate, Advanced

Language/Tools: Linux Kernel development (C), optional: FPGA synthesis flow

Mentors: Stefan Wallentowitz, Philipp Wagner, Stafford Horne

OpenRISC GCC Port

OpenRISC is a processor architecture which is simple, yet widely used for soft cores and in deeply embedded scenarios. For a long time a port of the GCC compiler for OpenRISC has been in existence. However, this port cannot be submitted upstream due to FSF copyright assignment issues. Hence a clean room rewrite of the basic compiler architecture is required.

This work gives you great insight into the backend of a compiler and the OpenRISC Instruction Set Architecture. To be successful in this project, you should have an understanding of compilers and be able to work in a very structured way. You will be mentored by an OpenRISC expert, with the help of a seasoned GCC developer.

Skill Level: Advanced

Language/Tools: C, Compilers

Mentor: Stafford Horne

GHDL profiling support

GHDL is a compiled VHDL simulator. This project would aim to profile and/or implement a profiler to improve speed of simulation and/or to detect hotspot in user design

Skill level: Intermediate

Language/Tools: ADA, VHDL

Mentor: Tristan Gingold

GHDL: Improve LLVM backend

This project would aim to improve the LLVM backend of GHDL, a compiled VHDL simulator.

Currently, llvm can be used as a backend, but this could be improved. In particular, there is no generation of debugging information, the C++ API should be used instead of the C API, and there was no real try to find the best order of optimization passes.

Skill level: Intermediate

Language/Tools: ADA, C++, VHDL

Mentor: Tristan Gingold

GHDL: Support 64-bit with mcode on Windows

GHDL has its own code generator (named mcode), which is light and fast but doesn’t try to optimize. Although x86-64 is supported on linux, win64 isn’t.

Skill level: Intermediate

Language/Tools: ADA, C++, VHDL

Mentor: Tristan Gingold

GPIO Switchbox

Details: FPGA boards are equipped with manyfold IO interfaces. Most of those IO interfaces are general purpose IO (GPIO) that are connected to standardized headers. One example are PMOD connectors or the Arduino shield connector.

If you build a system-on-chip nowadays the IO functions, such as GPIO, UART or SPI, are hard-wired to pins on those headers. Connecting for example other PMODs required re-synthesizing the design. I have started PMOD switchbox to allow the programmer to dynamically configure the IO functions at runtime. The goal is to generalize this approach and build an interesting prototype.

Skill level: Beginner, Intermediate

Language/Tool: Verilog

Mentor: Stefan Wallentowitz

FuseSoC Generators

Goal: Create a way to run external tools that generates files on-the-fly in a FuseSoC workflow

Details: When running a simulation or a synthesis flow it is often desirable to generate some files on-the-fly. This can be for generating RTL files from vendor-specific IP core descriptions, IP-XACT files or create an ELF file from C sources.

While FuseSoC already has support for calling external scripts at different parts of the toolflow, and many EDA tools have built in support for TCL scripting, it lacks a deeper integration into FuseSoC that is needed to efficiently perform some of these tasks.

The task includes both support for allowing cores to register themselves as generators as well as support for other cores to use them.

Examples of tasks that could make good use of generators

  • Xilinx CoreGen (replaces coregen provider)
  • Altera Qsys
  • Generate Wishbone interconnects (by running wb_intercon_gen in https://github.com/olofk/wb_intercon
  • Create verilog from Chisel/Migen/IP-Xact descriptions… etc

Skill level: Intermediate

Language/Tool: Python, basic knowledge of EDA tools

Mentor: Olof Kindgren

FuseSoC Provider for LibreCores API

Details: Currently, FuseSoC fetches the core description files from local copies of repositories. The goal of this project is to fetch the core descriptions from the LibreCores website instead.

In a first step, you must first coordinate with our LibreCores team to define the additional requirements for the API. Then FuseSoC should be extended to provide core files from other sources and then implement the (recursive) fetching of core files.

Skill Level: Intermediate

Language/Tool: Python

Mentors: Stefan Wallentowitz, Olof Kindgren

Constrained randomised testing with coverage tracking in Cocotb

Goal: 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.

Details: 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

Skill Level: Intermediate

Language/Tools: Python

Mentor: Chris Higgs

PCIe bus functional model (BFM) using Cocotb

Goal: Develop a PCIe bus functional model (BFM) in Python for use with hardware simulations via Cocotb.

Details: PCIe is a high-speed serial bus interface and is widely used in modern computing systems. It is commonly used to interface FPGAs and other ASICs to PCs or other host controller devices. The protocol is reasonably complex, and a good open source bus functional model (BFM) which can be used in hardware simulations of these FPGAs or ASICs is lacking.

Cocotb is a set of libraries which provide access to hardware simulation and essentially allows us to write testbench code in Python.

Given the complexity of the PCIe standard and the ease-of-use of Python with the cocotb libraries to interface to a hardware simulation, it is proposed that a PCIe BFM developed in cocotb-compatible Python.

The work in this project would be to, at first, develop a reasonably simple PCIe-capable BFM and simple API for driving it. This would be tested against a simulation of a simple FPGA design using the Xilinx PCIe endpoint IP. Proving configuration writes and basic memory-mapped accesses can be performed will be sufficient to begin with.

This work will expose the student to the concepts of writing behavioural hardware models in Cocotb-compliant Python, and simulating them against real designs.

Skill Level: Intermediate to Advanced

Language/Tools: Python, Verilog

Mentors: Luke Darnell, Julius Baxter

Firrtlator Extensions

Details: firrtlator is a C++ library and tool to work on the Flexible Intermediate Representation for RTL (FIRRTL). The goal is to have a common library to interact with FIRRTL files, to integrate it with other tools, and to elaborate and optimize on the IR.

The project is at the beginning and I am open to all ideas around it. Here are some suggetions:

  • Integration with Verilator
  • Integration with the YoSys RTL-IR
  • Automatic injection of debug probes
  • Injection of logic to improve fault tolerance

Skill Level: Intermediate/Advanced

Mentor: Stefan Wallentowitz

A SystemVerilog Preprocessor

The SystemVerilog hardware description language has an extensive set of preprocessor features, and antlr has established itself as one of the best tools for building lexers and parsers. However it is not possible to parse SystemVerilog with antlr (or any other toolkit) unless the preprocessor directives have been processed. The preprocessor directives include among other things, parameterized macros, token pasting, conditionals, includes, etc. The goal of this project is to write this preprocessor. This will enable the creation, in the future, of antlr based SystemVerilog tools (code analyzers, compilers, simulators, synthesizers, etc).

Skill Level: Advanced

Language/Tools: Antlr, Java or C++, SystemVerilog

Mentors: Martin d’Anjou

LimeSDR: Integrate RISC-V Core

Details: LimeSDR is a flexible software-defined platform, that integrates an FPGA and a Lime Microsystems LMS7002M field-programmable RF transceiver, that is 2x2 MIMO and can be tuned anywhere from 100KHz to 3.8GHz. The FPGA is presently used to take care of things such as interfacing, digital down-conversion and tuning, and sample time-stamping.

Currently the Altera NIOS-II processor is used and in this project the goal is to replace it with an open source softcore implementing the RISC-V instruction set.

It consists of three parts:

  • Identify a suitable RISC-V processor core
  • Integrate the RISC-V processor core
  • Proof-of-concept of the updated software environment

Skill level: Intermediate

Language/Tools: HDL, FPGA synthesis

Mentor: Andrew Back

LimeSDR: Switch to a FuseSoC-based Development Flow

Details: The LimeSDR-USB platform integrates an FPGA that contains a System-on-Chip and DSP. The System-on-Chip employs a softcore for interfacing with the programming interface of RF transceiver, while hardware blocks take care of streaming, DSP and sample time-stamping.

With the integration of more and more blocks and a varying number of FPGA implementations for different users, a structured way of managing the dependencies and composition of hardware blocks is needed. For that we would like to provide a FuseSoC-based environment.

The goal of this project is to structure the repositories and add the infrastructure files for FuseSoC, along with a proof-of-concept demonstration and a tutorial.

Skill level: Beginner, Intermediate

Language/Tools: Understanding of HDL Synthesis

Mentor: Andrew Back

SystemC generator plugin for Kactus2

The project objective is to implement a SystemC generator for Kactus2, an open-source IP-XACT tool. IP-XACT captures designs and components independent of implementation language, but the ability to generate SystemC from IP-XACT would enable designers to leverage SystemC for simulation and verification of their IP.

Skill level: Intermediate

Language/Tools: C++/Qt, SystemC

Mentor: Esko Pekkarinen

3D browser for IP-XACT library for Kactus2

Browsing lists and tree structures in a graphical user interface is often tedious and relations between the objects may not be intuitive. Recently 3D visualization has become a rich alternative to traditional navigation. The project objective is to implement 3D browsing for IP-XACT objects in Kactus2 tool.

Skill level: Advanced

Language/Tools: 3D graphics, C++/Qt, User interface design

Mentor: Esko Pekkarinen

Integration of Cloud FPGAs with JavaScript

The incorporation of FPGAs into the cloud, opens many possibilities for new compute models. Hardware acceleration can become an integral and ubiquitous part of cloud computing. This project will support hardware acceleration of web applications by creating a communication channel between JavaScript, running in a web browser, and FPGA hardware, running on Amazon F1 FPGA instances. Imagine the possibilities for web pages with access to hardware-acceleration!

An F1 instance is a cloud CPU (host machine) with attached FPGAs. The host will be used as a web server. It will serve JavaScript content that includes functions that communicate via a websocket back to the server. The server/host will relay this data stream to the attached FPGA. A return path from FPGA to host to client will also be created.

Communication between host and FPGA will be implemented using a High-Level Synthesis flow in the CAOS framework, or using a Verilog-based flow with the F1 Hardware Development Kit plus C++ code in the F1 Software Development Kit. Successful completion will include demonstration of a round-trip communication from web client to FPGA and back. Success also requires clear documentation of the APIs and hardware configuration steps.

Skill level: Advanced

Languages/Tools: Amazon F1, CAOS (C/C++/OpenCL) or F1 HDK/SDK, JavaScript, HTML5, Web Sockets, Node.js.

Mentor: Steve Hoover

CPU Components in Transaction-Level Verilog

The RISC-V ISA has been the subject of massive interest in its short history due to its open nature. RISC-V CPU implementations in Chisel and BlueSpec Verilog show a degree of flexibility that is impractical in RTL design. But this is nothing compared to the flexibility that is possible with a new extension to Verilog, called Transaction-Level Verilog (TL-Verilog). Here’s your chance to develop open-source CPU components for the most-configurable, most-adaptable RISC-V CPU core generator, WARP-V. WARP-V can be configured as a low-power, slow-clock, single stage pipeline, a high-frequency seven-stage pipeline, or anywhere in between. You can even swap out the RISC-V ISA for a completely different ISA altogether.

You will implement CPU components, such as branch predictors, caches, etc., and add support for RISC-V ISA extensions to provide a rich set of selectable components for WARP-V. You will learn advanced design practices and CPU microarchitecture. TL-Verilog tutorials are available at makerchip.com. You’ll find WARP-V in the Makerchip IDE under “Tutorials”/”Examples”/”CPUs”.

Skill level: Intermediate/Advanced

Language/Tools: TL-Verilog, Makerchip IDE

Mentor: Steve Hoover

Formally Verify the WARP-V RISC-V CPU Core

WARP-V is an emerging open-source CPU core generator for RISC-V CPUs. It benefits from the flexibility of Transaction-Level Verilog (TL-Verilog) to support a wide range of implementations from simple CPUs for FPGAs through high-frequency ASIC implementations. You’ll use open-source formal verification tools to prove that WARP-V implements the RISC-V specification. You’ll debug and complete missing elements of the microarchitecture. Find WARP-V in the Makerchip IDE under “Tutorials”/”Examples”/”CPUs”.

Skill level: Intermediate

Language/Tools: Yosys, riscv-formal, TL-Verilog, Makerchip IDE

Mentor: Steve Hoover

Develop a Transaction-Level Verilog Component Library

Transaction-Level Verilog (TL-Verilog) is an emerging language extension to SystemVerilog that is introducing a number of game-changing capabilities. Among them is the ability to define flexible reusable components. For this project, you will develop a library of compatible components such as FIFOs, queues, arbiters, elastic buffers, etc. that could form the basis of a new era of hardware design. You will demonstrate the ease of composing these components into sophisticated transaction flows and quickly implementing what are currently considered to be complex designs, such as a complete on-chip network!

Skill level: Intermediate/Advanced

Language/Tools: TL-Verilog, Makerchip IDE

Mentor: Steve Hoover

Chisel learning journey

Implementing support for standard debugger(gdb), fpu and caches in sodor and teach important chisel concepts through it. All of this while keeping documentation as first priority

Skill Level: Intermediate

Language: Chisel, Scala and Verilog

Mentor: Stefan Wallentowitz

Capturing the dynamic rocket

Find and implement some effective way to capture rocket-chip ecosystem which includes rocket, boom, uncore and hardfloat into micro-architecture diagrams/schematics (possibly using draw.io) and what makes it tough is parameterizable nature of this ecosystem, which also needs to be captured

Skill Level: Advanced

Language: Chisel, Scala, Verilog

Mentor: Stefan Wallentowitz

Yosys Test Suite

Yosys is a framework for HDL synthesis and formal verification. This project is about extending the Yosys test suite to comprehensively cover the language constructs in Verilog 2001 and to comprehensively cover all script commands available in Yosys, and integrate the test suite with a CI infrastructure.

A comprehensive test suite can help development of Yosys by providing a high degree of confidence that pull requests that pass the test suite do not introduce new bugs. Therefore this project is vital to opening Yosys to a larger community of open source developpers.

Skill Level: Intermediate

Language/Tools: Yosys, Verilog, GNU Make, Python

Mentor: Clifford Wolf

SystemVerilog Support in Yosys

YoSys is a framework for Verilog RTL synthesis. It does not support SystemVerilog so far. This requires extension of the frontend of Yosys and a good understanding of how to map System Verilog constructs to the IR.

Skill Level: Advanced

Language: C

Mentor: Tim Ansell

Common EDA tool output format

Proprietary and open source EDA synthesis and implementation tools all have different output formats for warnings, error and synthesis results. A translator tool that converts all outputs to a common format is of high value for integration with other tool flows or continuous integration. The goals are a definition of the format in coordination with the community and then the implementation of this tool.

Skill Level: Beginner, Intermediate

Language: tbd

Mentor: Tim Ansell, Stefan Wallentowitz

Solar Cell Simulation

DEVSIM is simulation software which uses the finite volume method to solve electrical behavior for semiconductor devices on a microscopic level. It solves partial differential equations on a mesh. The Python interface allows users to specify their own equations.

DEVSIM is a simulator capable of solving for the behavior of individual semiconductor devices due to external voltage boundary conditions. To simulate a solar cell, it is necessary to simulate the propagation of light into the structure and the resulting electrical current. This would require:

  1. Reading in the existing structure from an external mesh tool tailored toward solar cell structure generation.

  2. Modeling the propagation of light through the semiconductor structure using one of these methods:

  • Ray tracing
  • Transfer matrix method
  • Beam propagation method
  • Finite difference time domain
  1. Implementing the equations coupling light intensity and the generation of current in the semiconductor.

  2. Simulate figures of merit: Isc, Voc, Jsc, Pmax, Power curve, Fill Factor and IV curve.

Based on the scope of this project, a student should be able to tackle reading in the structure, and solving for the propagation of light using one of the listed methods. In an initial implementation this would be for 1D and 2D structures. Based on the student’s interest, they may choose to tackle the problem for a 3D structure. Alternatively, they may choose to address the models for generation of current from light. The system would then be well suited for analyzing the device performance.

Skill Level: Advanced

Languages/Tools: C++, Python

Mentor: Juan Sanchez

Digital filter blocks

This project includes developing various digital filters with MyHDL and integrating the filter-blocks with PyFDA. The filter-blocks will be added to the filter-blocks repository.  The student is encourage to start with simple digital filters and use the tools to verify and demonstrate the performance for the filters. 

After a basic filter or two is completed it is expected the student will complete a more complex digital filter, examples: systolic structure, non-linear estimation filters, modular shared-resource structures, etc.

Skill level: Intermediate, Advanced

language/Tool: MyHDL

Mentor: Christopher L. Felton, Christian Münker

Interface driven design, Ethernet core

This project involves extending the Ethernet core that was developed in a previous GSoC to be an “interface driven” design.  The core is to be updated and extended with the interfaces provided in the rhea or crux packages. The following example illustrates the idea of the interface driven design - the details will be developed with the student.

def example(...)
    glbl = Global()         # clock and resets
    ethcfg = EthConfig()    # contains all the setup info
    ethc = EthController()  # main data interface to the eth
 
    # In this example, there are three blocks that are interfacing
    # with the Ethernet controller.
    b1 = ethc.get_branch()
    pkt1_inst = packet_type_a_handler(...)
 
    b2 = ethc.get_branch()
    pkt2_inst = packet_type_b_handler(...)

    b3 = ethc.get_branch()
    pkt3_inst = packet_type_c_handler(...)
 
    eth_inst = eth_core(glbl, ethc, ethcfg)
    # get the arbitration logic, handles
    arb_inst = ethc.arbitrate()  

It is desirable, but not required, that the student have a development board with an tri-mode (10/100/1000) Ethernet phy.

Skill level: Intermediate, Advanced

language/Tool: MyHDL

Mentor: Christopher L. Felton , Tom Dillon

Rocket versus Ariane RISC-V “Deathmatch” and Analysis

The RISC-V ISA will transform the world. Recently, two Linux-capable cores, one produced by SiFive/Berkeley, and one produced by ETH Zurich have been produced. The goal of this project is to do a detailed performance analysis of these two cores, and provide recommendations and best practices back to those efforts for the path to increase performance.

This group will be performed in conjunction with the Bespoke Silicon Group team at U. Washington (which architected the 16nm 511-core RISC-V core processor, Celerity.)

Skills: Knowledgable of both in-order and out-order processor architecture. Hardware design skills.

Language/Tools: SystemVerilog, Comfortable with Unix Programming and Scripting

Mentor: Prof. Michael Taylor and the Basejump Team

BaseJump RISC-V Manycore CUDA Design

The RISC-V ISA will transform the world. Recently, U. Washington taped out an open source RISC-V manycore processor with 496 cores that hits 500 Billion RISC-V instructions per second in one chip.

We have silicon up and running in our lab and will be taping out a second generation based on results from the first generation. Our goal is to make this the defacto open source GPGPU design. (See bjump.org/manycore.) You will work to help design the CUDA-like programming environment and make recommends for future versions of the architecture.

Skills: Knowledgeable of CUDA programming; Compiler or PL interests a plus.

Language/Tools: C, Comfortable with Unix Programming and Scripting

Mentor: Prof. Michael Taylor and the Basejump Team

BaseJump STL Hacker

We are prototyping new features for the SystemVerilog language, including the equivalent of the C++ Standard Template Library, which we intend to deploy into the SystemVerilog standard.

Our initial results show that HW productivity is dramatically improved with this library. An initial version of this is up and running, and we are looking for folks with an outside perspective to take it to the next level.

Skills: Hardware Design Experience, Interest in HDL (hardware description language) design

Language/Tools: SystemVerilog

Mentor: Prof. Michael Taylor and the Basejump Team

VHDL Frontend for Yosys

VHDL is one of the “major” hardware description languages alongside Verilog. Yosys is an open-source HDL synthesis tool, and it currently only supports Verilog (it supports VHDL only via the proprietary Verific frontend). The goal of this project is to write an open-source VHDL frontend for Yosys.

There are currently a number of open-source tools that support VHDL that can be leveraged as a base. For example, there is GHDL, a VHDL simulation tool written in Ada; NVC, a VHDL simulation tool written in C; vhdlpp, a VHDL frontend for the Icarus Verilog simulator; and Robert Ou’s stalled project, a VHDL parser written in mostly Bison and Rust. You may choose to start with any one of these existing projects or any other projects you may discover (starting from scratch is likely to be too difficult given the time limitations).

Skill Level: Advanced

Language/Tools: VHDL, C, C++, Ada(?), Flex/Bison(?), Rust(?)

Mentor: Robert Ou

Framework for Mixed-Language Simulation

There are a number of open-source HDL simulators currently available. For example, Icarus Verilog and Verilator support Verilog, and GHDL, NVC, and Icarus Verilog (via vhdlpp) support VHDL. However, most of these support only one of VHDL or Verilog (Icarus supports both, but its VHDL support is incomplete), and it is very difficult to simulate projects using a mix of both languages. The goal of this project is to design some method for doing mixed-language HDL simulation with open-source tools that has good support for the features of all of the HDLs.

Some possible solutions to this problem include:

  • Write a program that uses the procedural interfaces (i.e. VPI, VHPI) of each simulator and somehow “glue” them together (one such hack that does so is here)
  • Lower all of the HDLs into a common intermediate representation and then write/modify a simulator to simulate that

Some other things that you should consider include support for HDLs other than Verilog and VHDL (e.g. MyHDL, Migen, Chisel, etc.) and cosimulation (e.g. simulation including external software, simulation with hardware in the loop).

Skill Level: Intermediate/Advanced

Language/Tools: VHDL, Verilog, C, C++

Mentor: Robert Ou

Recent activity

26 Feb 2019 » Become a GSoC student 2019!
We are happy and honored to be part of the Google Summer of Code (GSoC) again this year. We are..... more

24 Feb 2019 » Announcing Latch-Up in Portland, Oregon
The FOSSi Foundation is proud to announce Latch-Up (latchup.io), a conference dedicated to free and open source silicon to be..... more

07 Nov 2018 » Videos and slides from ORConf 2018 now available!
ORConf 2018 was a great success, with many fantastic talks, hallway discussions, and valuable insights into other projects. If you..... more

12 Oct 2018 » El Correo Libre Issue 8
The FOSSi Foundation’s monthly roundup of all things Open Source semiconductor design, including a wrap-up of ORConf 2018, can be..... more

15 Aug 2018 » El Correo Libre Issue 6
You can find the new issue of our monthly newsletter here: El Correo Libre Issue 6 – LibreCores – Medium..... more

The Free and Open Source Silicon Foundation C.i.C.
Unit C5, Tenterfields Business Park, Halifax, England, HX2 6EQ

info@fossi-foundation.org