Talk Title:
High Productivity & High Performance Through Computer-Aided Programming

Abstract:
Due to the end of Dennard scaling, parallelism has become the only path forward for maintaining performance gains from Moore's Law. At the same time, the proliferation of accelerators, hierarchical architectures, GPUs, and mobile devices has led to a multiplicative effect on the amount of time required for programming these systems to obtain high performance. Furthermore, maintainability has suffered since the usual optimization techniques require hand-rewriting code in ways that mix semantic meaning with optimizations. This talk presents research towards a path forward using domain-specific embedded languages and compilers, program synthesis, and automated optimization/learning techniques together to build new systems that enable programmers to express high-level intent and obtain portable high performance, while increasing both productivity and maintainability. With these new programming systems, we use computers to make high performance programming simpler and easier.

Speaker Bio:
Shoaib Kamil is a research scientist at the Computer Science and Artificial Intelligence Laboratory (CSAIL) at the Massachusetts Institute of Technology, working with Profs. Saman Amarasinghe and Armando Solar-Lezama. His research spans the areas of high performance programming and computing, domain specific languages and compilers, and program synthesis. Prior to MIT, he obtained his PhD from the University of California at Berkeley where his research became the centerpiece of two large multi-PI projects, including the Parallel Computing Laboratory, funded by Intel and Microsoft.