Workshop goal
Increasingly, high-performance applications are written using various high-level abstractions, from libraries to domain-specific languages, and with each passing year, new abstractions and models are being proposed. These abstractions often possess specific semantic properties, either domain-specific (such as mathematical properties of a domain's operations) or general (such as locality or layout information in a data-structure library). Nevertheless, by taking advantage of these abstractions or semantic properties, applications can be restructured, computations can be rescheduled, algorithms can be refined and data layouts can be tuned to improve performance.
The goal of this workshop is to bring together researchers who are investigating how to take advantage of high level semantic information, abstractions and domain-specific knowledge to write, optimize and transform high-performance applications.