HLS Directives Design Space Exploration Using Genetic Algorithms

This project features an optimizer designed to automatically conduct High-Level Synthesis (HLS) directives Design Space Exploration (DSE) through genetic algorithms.

Supported HLS Directives:

• Loop Pipeline
• Loop Unroll
• Array Partition

Optimizer Inputs:

1. Synthesizable C/C++ source code with labels at each action point (e.g., arrays, loops). Action points are labeled sequentially (L1, L2, etc.).
2. kernel_info.txt that includes:
   • a) The top-level function name
   • b) Details for each action point, including loop trip counts for loop action points and array names with dimension sizes for array action points.

Optimizer Outputs:

1. Pareto-optimal kernel source codes
2. info.csv detailing trade-offs among Pareto-optimal kernel source codes
3. A database containing examined directive configurations with their respective latencies and resource usage (BRAM%, DSP%, LUT%, and FF%)
4. APP_NAME.json providing statistics for the database

Sample inputs and outputs can be found in the dataset and databases directories.

Getting Started

Follow these instructions to get a copy of the project on your local machine.

Prerequisites

This project has been tested on Ubuntu 18.04.6 LTS with Python 3.6.9 and Vitis 2021.1 suite. Additionally, the following libraries are required:

• pymoo (v0.5.0)
• sqlitedict (v2.0.0)
• psutil (v5.9.0)

Install these dependencies using:

python3 -m pip install -r requirements.txt

Running the Project

After installing the necessary software from the Prerequisites section, clone this repository to your local machine.

Execute the HLS Directives Design Space Exploration:

./exec.sh run APP_NAME EXTENSION

Example:

./exec.sh run rodinia-knn-1-tiling .cpp