Code Size Reduction

Angha Source Code

Original Source Code

Reduced Source Code

Representations

Inst2Vec

LLVM intermediate representation (Angha 30k)

LLVM intermediate representation (Angha + MiBench)

Milepost Static Feature

Program graphs for machine learning

ProGraML -> NetworkX

Training Data

The list of benchmarks

Training for binary size

Euclidean distances

Training for number of instructions LLVM

The good sequences