Projects
Habitat

Authors: Geoffrey Yu, Yubo Gao, Pavel Golikov, Gennady Pekhimenko

Published at USENIX ATC 2021, Habitat is a new practical technique to help users make informed and cost-efficient GPU selections: make performance predictions with the help of a GPU that the user already has.
Skyline

Authors: Geoffrey Yu, Tovi Grossman, Gennady Pekhimenko

Published at UIST 2020, Skyline is a new interactive tool for DNN training that supports in-editor computational performance profiling, visualization, and debugging.
LifeStream

Authors: Anand Jayarajan, Kimberly Hau, Andrew Goodwin, Gennady Pekhimenko

Published at ASPLOS 2021, LifeStream is a high-performance stream processing engine for physiological data.
HFTA

Authors: Shang Wang, Peiming Yang, Yuxuan Zheng, Xin Li, Gennady Pekhimenko

Published at MLSys 2021, Horizontally Fused Training Array is a new deep learning framework extension library that horizontally fuses the models from different repetitive jobs deeply down to operators and then trains them simultaneously on a shared accelerator.
RL-Scope

Authors: James Gleeson, Srivatsan Krishnan, Moshe Gabel, Vijay Janapa Reddi, Eyal de Lara, Gennady Pekhimenko

Published at MLSys 2021, RL-Scope is a cross-stack profiler that scopes low-level CPU/GPU resource usage to high-level algorithmic operations and provides accurate insights by correcting for profiling overhead.
IOS

Authors: Yaoyao Ding, Ligeng Zhu, Zhihao Jia, Gennady Pekhimenko, Song Han

Published at MLSys 2021, Inter- Operator Scheduler is an inter-operator scheduler that accelerates CNN inference by combining intra- and inter-operator parallelism.