This compiler converts LLVM IR to SWPP assembly.
NOTE: Please don't just fork this and use it as your team repository! It will make your team repository public (visible to other teams). Instead, create an empty private repository, copy the contents of this repo to yours.
Whenever there is a change in this repo, you can cherry-pick the new commits. Relevant links: here, here.
We have taken a powerful register allocation algorithm from the last competition winner team (special thanks to swpp202001 team number 1) and patched it to support vector instructions(vload and vstore). Our purpose was to give you a solid backend to focus more on the IR level optimizations.
To compile this project, you'll need to clone & build LLVM 12.0 first. Please follow README.md from https://github.com/aqjune/llvmscript using llvm-12.0.json. If you did the assignments, you will already have built this.
After LLVM 12.0 is successfully built, please run:
./configure.sh <LLVM bin dir (ex: ~/llvm-12.0-releaseassert/bin)>
make
You should implement test part inside Makefile.template to support the underlying test command:
make test
Compile LLVM IR input.ll into an assembly a.s using this command:
bin/sf-compiler input.ll a.s
For more information, run bin/sf-compiler -h.