HIPERFIT Talk: Functional Array Streams

Presenter: Frederik Meisner Madsen, PhD student, DIKU/University of Copenhagen

Time: Friday August 28, 2015, 13-14

Place: The APL Meeting Room (DIKU, HCØ, Universitetsparken 5; room 01-0-029)

Abstract:

Regular array languages for high performance computing based on aggregate operations provide a convenient parallel programming model, which enables the generation of efficient code for SIMD architectures, such as GPUs. However, the data sets that can be processed with current implementations are severely constrained by the limited amount of main memory available in these architectures.

In this paper, we propose an extension of the embedded array language Accelerate with a notion of sequences, resulting in a two level hierarchy which allows the programmer to specify a partitioning strategy which facilitates automatic resource allocation. Depending on the available memory, the runtime system processes the overall data set in streams of chunks appropriate to the hardware parameters.

In this paper, we present the language design for the sequence operations, as well as the compilation and runtime support, and demonstrate with a set of benchmarks the feasibility of this approach.

Biography

Frederik Meisner Madsen is a PhD student at the HIPERFIT research center where he is involved in programming language research. Particular areas of interest include functional array languages, nested data-parallelism and streaming.

Host: Martin Elsman



Published

28 August 2015

Tags