Repository for the course PPN (Projet de Programmation Numérique): AI-ACCE (AI based Assembly Code Cycles Estimator)

Link : https://academic.liparad.uvsq.fr/m1chps/ppn/projets.html#ai-acce-ai-based-assembly-code-cycles-estimator

Current status of the project

Done/To do:

• Set up the project
• Add parsing of the assembly code
• Get data to train the AI model
• Implement the AI model
• Improve accuracy of the AI model
• Improve speed of the AI model

Project is now finished for the course. We don't plan on touching it anymore.

Dependencies

• C++17
• CMake
• GTest
• clang-format and clang-tidy
• Doxygen (if you want to generate the documentation)
• nanobench (already included in the project as a submodule)
• eigen

How to run the project

• Clone the repository with submodules: `git clone --recurse-submodules
• Build it with CMake:
  • mkdir build
  • cd build
  • cmake ..
  • make
• Now you can run the tests with make check
• You can extract the data with $/build/> ./extract_data
• You can run the AI model with $/> ./main

Project structure

├── benchmarks              # Benchmarking the neural network
│   ├── bench.cpp           # Contains the code to benchmark the neural network
├── codelets                # Contains the kernels to generate the data
│   └── op.cpp
├── bin                     # Contains the compiled files (including the tests and benchmarks)
├── dataset                 # Contains the training data
│   ├── bench_bins          # Contains the binaries of the benchmarks along with the cycles count
│   ├── bench_bins_small    # Contains a small subset of the benchmarks binaries for testing
├── docs                    # Contains the documentation
├── nanobench               # Contains the nanobench submodule
├── experiments             # Contains several experiments using pre-made frameworks we did not related to the main project
├── include                 # Contains the header files
│   └── x.hpp
├── rapport                 # Contains the report of the project
├── results                 # Contains the results of the experiments
├── src                     # Contains the source code
│   └── x.cpp
├── tests                   # Contains the tests source code
│   └── test_x.cpp
├── CMakeLists.txt          # To compile the project
├── default.nix             # Nix file to create a development environment
├── Doxyfile                # Doxygen configuration file
└── .clang files            # Contains the clang files for formatting and linting like .clang-format, .clang-tidy and .clangd

Move the pre-commit.sh file to .git/hooks/pre-commit to make the project automatically format, test and generate the documentation when committing.