esalcort / performance_estimation

Performance Estimation

Performance estimation exercises using machine learning

Contents

• gemm_runtime_estimation/
  • gemm_nn.c: A dummy matrix multiply code with variable input sizes M, N, and K
  • sample.csv: A set of profiling results after running gemm_nn with different inputs
  • estimation_models.ipynb: A jupyter notebook with examples to train ML models that predict data in sample.csv
• cnn_mapping/
  • monte_carlo_tree_search.py: A minimal implementation of MCTS imported from [1]
  • tictactoe.py: An example application of MCTS imported from [1]
  • simulate_yolov3: A class that estimates computation and communication runtime of a yolov3 tiny CNN running on two processors
  • map_yolo.py: A class that implements the MCTS Node class for the simulate_yolov3 file

References

[1] Luke Harold Miles, monte_carlo_tree_search.py, Github Gist, https://gist.github.com/qpwo/c538c6f73727e254fdc7fab81024f6e1

Contact

If you have questions or suggestions about this repository, or a problem running examples, you can contact Susy at [email protected]