Qameleon: Hardware/software Co-operative Automated Tuning for Heterogeneous Architectures

点击翻译按钮获取中文摘要

"This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5)."The push toward heterogeneous architectures to increase performance, while reducing energy consumption creates considerable challenges for software development. For example, programmers must make non-trivial decisions about when to use special accelerators vs. powerful core CPUs and also become steeped in complex architectural details to tune effectively. The goal of this research project is to alleviate these challenges using a novel framework that enables a wide-range of computations to be expressed at a high-level and subsequently tuned automatically for the underlying heterogeneous platform. More specifically, the PIs propose Qameleon, a new programming environment that can cooperatively tune the program and the hardware configuration automatically and continuously using statistical machine learning techniques. The proposed work will be the first in GPU programming to consider adaptively partitioning a computation on a heterogeneous platform at run-time. This work will also improve understanding of the trade-offs among programming features, architectural support, performance, and power in heterogeneous architectures. The research will also develop several metrics to characterize the application based on the outcome of the statistical modeling. The proposed research brings together cross-disciplinary techniques?from architectures, compilers, machine learning, and applications ? and researchers from both academia and industry to build new common programming interfaces that can hide the complexity of heterogeneous architectures from the programmers, while still providing high-performance and energy-efficient execution. The Qameleon programming environment will be designed to teach at the undergraduate level by incorporating research results into new undergraduate courses aimed at both computer scientists and domain scientists alike.