Skip to content

system_setup.md

Latest commit

 

History

History
167 lines (128 loc) · 7.84 KB

system_setup.md

File metadata and controls

167 lines (128 loc) · 7.84 KB

System Setup

Chisel

Chisel is a Domain Specific Hardware Description Language developed at UC Berkeley. It is embedded in Scala which uses the sbt build tool and package manager to manage dependencies.

Install sbt

sbt automatically downloads everything you need, including the exact version of sbt itself that you need for the project. Thus the only thing you need for Chisel and Scala is to installsbt. Follow the instructions in the official installation guide.

Install Verilator

In order to simulate the low level Verilog code generated by Chisel you need the open source circuit simulator verilator. At least version 3.886 is required. On Fedora 27, you can install version 3.890 from the default repositories. On a recent Ubuntu 17.04 or Ubuntu 17.10 version 3.9 is available from the default repositories.

If your distribution does not offer a recent version of verilator, you can compile it yourself by following the Chisel Readme.

Test your Installation

You can test your installation by running the chisel tests:

  1. change to the chisel directory in this repository: > cd chisel
  2. launch sbt: > sbt
  3. once sbt is booted, run the tests: sbt:queue> test
  4. The first test (with firrtl) should succeed if you have installed sbt and Chisel is able to run on your system correctly.
  5. The second test (with verilator) will succeed if your verilator installation is working.

Vivado

We recommend using version 2017.3 which can be downloaded directly from Xilinx. Use the All OS installer Single-File Download which has proven to be reliable.

Install Vivado

In the explaination below we assume that you want to install Vivado into a user writable directory instead of a system directory. We use /home/kevin/xilinx_test as the example path. If you want to install to e.g. /opt, you need to launch xsetup with sudo.

  1. extract the Vivado files: tar -xf Xilinx_Vivado_SDK_2017.3_1005_1.tar.gz
  2. change into the extracted directory and launch ./xsetup
  3. click next
  4. accept all three license agreements and click next
  5. select Vivado HL WebPack (this is the free - as in beer - version) and click next
  6. we only need Vivado and Zynq-7000 support: vivado installation selection
  7. click next
  8. select the installation directory. We recommend to deselect all shortcut options on the right: installation directory
  9. review your selection and install
  10. wait until you see the Installation completed successfully. promt.

Install Board File

By default, Vivado does not know about the PYNQ board that we are using. Luckily, digilent provides a board description file for the very similar arty-z7-20 board.

  1. Checkout the digilent board files repository: git clone https://github.com/Digilent/vivado-boards.git Alternatively you can download and extract the ZIP file: wget https://github.com/Digilent/vivado-boards/archive/master.zip
  2. Copy the arty-z7-20 board file into your Vivado installation, e.g.: cp new/board_files/arty-z7-20 /home/kevin/xilinx_test/Vivado/2017.3/data/boards/board_files -r

Request and Install a Vivado License

To be able to use Vivado, a free (as in beer) license needs to be requested using a Xilinx account and then installed.

  1. Launch Vivado: source /home/kevin/xilinx_test/Vivado/2017.3/settings64.sh && vivado
  2. Launch the License Manager by clicking on Help -> Manage License
  3. Got to the Optain License tab, select Get Free WebPack [...] and click on Connect Now optain license
  4. If this does not work, try Save Link As and manually navigate to the website.
  5. Follow the instructions on the Xilinx website to optain a Xilinx.lic file.
  6. Go to the Load License tab and click on Copy License.
  7. Navigate to the Xilinx.lic file that you downloaded in step 5 and select it.
  8. After the success promt, you can exit the License Manager and then exit the Vivado GUI.

Test your Installation

This requires that you have installed and tested Chisel.

  1. change to the chisel directory in this repository: > cd chisel
  2. launch sbt: > sbt
  3. once sbt is booted, generate the low level Verilog in chisel/ip: sbt:queue> run
  4. quit sbt
  5. change to the root directory in this repository: > cd ..
  6. load Vivado: source /home/kevin/xilinx_test/Vivado/2017.3/settings64.sh
  7. synthesize the example design and generate the bitstream: make
  8. if this succeeds, your Vivado and Chisel installation are working

Rust

In order to control the circuits on the FPGA, we will use the modern systems programming language Rust. Compared to C, Rust offers modern programming features such as traits, pattern matching, a module system, a collections library and much more. Whereas a lot of programs written in C or C++ suffer from memory errors, the Rust compiler can prove at compile time, that our code is free of these errors. Rust is a better fit for system programming compared to Python or Java, since it compiled into a single static binary without any complicated runtime requirements (such as a garbage collector).

RustUp

The recommended way of installing rust is to use the rustup tool. Just follow their installation instructions.

Cross Compile Toolchain

We will be targeting the ARMv7 core on the Zynq chip to run our program on. For this purpose we need to install a rust toolchain that compile to this target:

> rustup target add armv7-unknown-linux-gnueabihf

This installs the rust compiler for the armv7 target, however, we still need a linker. For this purpose you need to install the gcc cross compile toolchain for this target.

On Fedora you can find a compatible toolchain on copr. On Ubuntu you can just install the gcc-arm-linux-gnueabihf package.

More details are explained in this readme, but please be aware that creating a .cargo configuration is not needed since we include a local version in this repository.

Test your Installation

To test whether your Rust installation can cross compile our code:

  1. change to the control directory in this repository: > cd control
  2. build the application: > cargo build --target=armv7-unknown-linux-gnueabihf
  3. make a release build: > cargo build --target=armv7-unknown-linux-gnueabihf --release
  4. if there are no error, you are ready to test the application on the FPGA

Chisel Bootcamp Jupyter Notebook

In order to experiment with and learn Chisel without an FPGA at hand, you can checkout the Jupyter notebooks that were created by folks at Berkeley as interactive teaching material. Just follow their installation guide.
⚡