ATMOSPHERE was presented at the 30th International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 2018), from 24th till 27th September, Lyon (France). SBAC-PAD is an international symposium, started in 1987, which has continuously presented an overview of new developments, applications, and trends in parallel and distributed computing technologies. SBAC-PAD is open for faculty members, researchers, specialists and graduate students around the world.
ATMOSPHERE project presented a paper that proposes a new approach to model the execution time for training convolutional neural networks (CNNs) deployed on GPUs
The paper, entitled "Performance Prediction of GPU-based Deep Learning Applications", demonstrate that ATMOSPHERE approach is generally applicable to a variety of CNN models and different types of GPUs with high accuracy, aiming at the preliminary design phases for system sizing. More info available on the full paper