Emmet Murphy won 1st place in the Machine Learning Contest for Chip Design with High-Level Synthesis
Emmet Murphy, doctoral student in the DYNAMO group headed by Prof. Lana Josipović, won the first place in the Machine Learning Contest for Chip Design with High-Level Synthesis (Prediction task), organized by AMD and UCLA.
In the contest, participants competed in predicting the area and performance of circuits that can be generated from C++ code descriptions. Emmet used Balor, his open-source code evaluator based on graph neural networks, to perform these predictions with the highest accuracy.
More on the competition: external page Machine Learning Contest for Chip Design with High-Level Synthesis
Link to Balor, Emmet's predictor that won the contest: external page https://github.com/emmet-murphy/balor
Link to the paper describing Balor: Balor: HLS Source Code Evaluator Based on Custom Graphs and Hierarchical GNNs
Congratulations!