High-valued domain applications in areas such as medicine, biology, physics, engineering, and social phenomena demand both fast innovation and high execution speed and require productive development environments for domain experts who may not be computer science experts. This workshop brings together leading researchers in architecture, compilers and programming languages, and domain experts to discuss and debate potential approaches to accelerating progress in such high-valued domains with an emphasis on developing strategies for exploiting machine learning, including strategies for accelerating learning algorithms through parallelism. The goal is to stimulate an in-depth discussion of the potential benefits of joint architecture and compiler approaches. The workshop will promote broadening participation by including speakers from groups underrepresented in computing and early career researchers.
The workshop will produce a report providing recommendations on: joint compiler/language and architecture approaches, compiler/language support enabling more aggressive hardware capabilities, architecture support enabling more effective compilers, and applications whose development process could benefit by these advances, The report will identify research opportunities in the interaction between developers, and architecture and language/compiler researchers to enable productive domain application development and highly efficient and scalable implementation on heterogeneous computing systems. The report will outline promising approaches and the research required for these approaches to become usable by the domain application developers.
