Visiting Lecture Series: Vivek Sarkar, "Software Challenges for Extreme Scale Systems or How to Play the End-Game for Moore's Law"
It is widely recognized that future generations of computer systems will be qualitatively different from current and past generations. Specifically, they will be built with thousands of homogeneous and heterogeneous processor cores per chip combined with heterogeneous memory hierarchies, their performance will be driven by parallelism, and will be constrained by fundamental energy and data movement limitations. They will also be subject to frequent faults and failures. Unlike previous generations of hardware evolution, these "extreme scale" systems will have a profound impact on future software. The software challenges are further compounded by the need to support application domains that have traditionally not had to worry much about parallelism in the past.
Addressing these challenges for extreme scale systems will require major advances in the core of computer science, including programming languages, compilers, runtime systems, and processor architectures. These advances are also critical for extending the end-game of Moore's Law, by contributing to exponential performance improvements in successive generations of computer systems without relying solely on exponential improvements in transistor density or cost.
In this talk, Vivek Sarkar will discuss approaches to addressing software challenges for extreme scale systems based on experiences gained in the Habanero ExtremeScale Software Research Laboratory at Rice University. Where the overall approach is based on introducing a set of unified primitives for structured parallelism and concurrency, which can be used to: 1) support a wide range of application patterns, 2) define a family of structured parallel programming models with conclusive semantic guarantees, and 3) obtain efficient implementations for single-node homogeneous/heterogeneous processor systems as well as multi-node cluster systems. Some of these primitives have already influenced industry standards for parallelism including the do across construct in OpenMP 4.5, the task blocks library for C++, and Java's Phaser library.
This work was done in collaboration with members of the Habanero Extreme Scale Software Research Laboratory at Rice University.
- Workflow Status: Published
- Created By: Devin Young
- Created: 03/17/2016
- Modified By: Fletcher Moore
- Modified: 04/13/2017