Quantum Engineering Colloquium


1. Erik Vermij,  TU Delft, The Netherlands -- 03-05-2017

2.Tom Hogevorst, TU Delft, The Netherlands -- 03-05-2017


1.    Near-memory computing architectures for data-intensive applications

In the field of data-analytics, companies and governments are piling up data and try to create value out of it, requiring novel data-intensive applications. In the field of high-performance computing, novel sparse methods are bounded by the memory system. The weakly-attached compute-centric coprocessors developed in the last decade offer no solution here.
By introducing strongly-attached, virtualized and coherent, coprocessors close to the main memory of a CPU, we introduce easy to use processing capabilities with memory access characteristics beneficial for this new class of applications. We discuss the architectural implications as well as results.

 2.   A Domain-Specific Language and Compiler for Computation-in-Memory Skeletons

 Computation-in-Memory (CiM) is a new computer architecture template based on the in-memory computing paradigm. CiM can solve the memory-wall problem of classical Von Neumann-based computer systems by exploiting application-specific computational and data-flow patterns with the capability of performing both storage and computations of emerging resistive RAM technologies (e.g., memristors). However, to efficiently explore and design such radically new application-specific CiM architectures, we require fundamentally new algorithm specification and compilation techniques. In this presentation, we introduce a domain-specific language to express not only the computational patterns of an algorithm but also its spatial characteristics. Furthermore, we design a compiler that is able to transform these patterns into highly-optimized CiM designs. Experiments demonstrate the functional correctness of the language and the compiler as well as an order of magnitude speedup improvement over a multicore system in both performance and energy costs.



Slides for the first talk is available here.

Slides for the second talk is available here.

CE Tweets