Awesome

Dahlia is a programming language for designing hardware accelerators. It provides abstractions that guarantee predictable hardware generation after type checking. For more details, see the demo and the documentation.

Fuse is the reference compiler for compiling Dahlia programs to various HLS backends. Vivado HLS is the only currently supported backend.

Set It Up

The compiler is written in Scala. To get things running, you will need a Java runtime, Scala itself, and sbt. Here's what you need to do:

• Get Java if you don't already have it.

  • For MAC OS: brew tap AdoptOpenJDK/openjdk && brew cask install adoptopenjdk8.
  • NOTE: The default adoptopenjdk (Java 13) version does not work with SBT.

• Install Scala and sbt. On macOS, use brew install scala sbt.

• Install Runt to run integration tests:

  cargo install runt \
    --version "$(grep ver runt.toml | awk '{print $3}' | tr -d "'")"
  

• Now you can compile the compiler by typing sbt assembly.

• Use sbt test to run the tests

• Use runt to run the integration tests from the repository root.

Compiler development

If you're working on the compiler, you probably want to use the sbt console instead (it's faster for repeated builds). Run sbt alone to get the console, where you can type commands like compile, test, and run [args].

Adding the prefix ~ (such as ~compile) makes sbt go into watch mode, i.e., it will re-run the command every time a dependency changes. Use ~assembly to continously update ./fuse or ~test to continously test the changes.

If you want to execute a sequence of sbt commands without starting sbt console, you can type sbt "; cmd1; cm2 ...". For example, sbt "; test; assembly" will run sbt test followed by sbt assembly.

Use It

Type sbt assembly to package up a fat jar for command-line use. The short fuse shell script here invokes the built jar to run the compiler. To compile a simple test, for example, run:

$ ./fuse ./file-tests/should-compile/matadd.fuse

The compiler produces HLS C source code on its standard output.

Source

Because of how Docusaurus is structured, the website is stored in the website/ directory and the documentation files are stored in docs/.

Building

We use github pages to deploy the page. Read the README under website/ for instructions.