Make the most out of your SIMD investments: Counter control flow divergence in compiled query pipelines

Harald Lang, Andreas Kipf, Linnea Passing, Peter Boncz, Thomas Neumann, Alfons Kemper

Publikation: Beitrag in Buch/Bericht/KonferenzbandKonferenzbeitragBegutachtung

9 Zitate (Scopus)

Abstract

Increasing single instruction multiple data (SIMD) capabilities in modern hardware allows for compiling efficient data-parallel query pipelines. This means GPU-alike challenges arise: control flow divergence causes underutilization of vector-processing units. In this paper, we present efficient algorithms for the AVX-512 architecture to address this issue. These algorithms allow for fine-grained assignment of new tuples to idle SIMD lanes. Furthermore, we present strategies for their integration with compiled query pipelines without introducing inefficient memory materializations. We evaluate our approach with a high-performance geospatial join query, which shows performance improvements of up to 35%.

OriginalspracheEnglisch
Titel14th International Workshop on Data Management on New Hardware, DaMoN 2018
Herausgeber (Verlag)Association for Computing Machinery, Inc
ISBN (elektronisch)9781450358538
DOIs
PublikationsstatusVeröffentlicht - 11 Juni 2018
Veranstaltung14th International Workshop on Data Management on New Hardware, DaMoN 2018 - Houston, USA/Vereinigte Staaten
Dauer: 11 Juni 2018 → …

Publikationsreihe

Name14th International Workshop on Data Management on New Hardware, DaMoN 2018

Konferenz

Konferenz14th International Workshop on Data Management on New Hardware, DaMoN 2018
Land/GebietUSA/Vereinigte Staaten
OrtHouston
Zeitraum11/06/18 → …

Fingerprint

Untersuchen Sie die Forschungsthemen von „Make the most out of your SIMD investments: Counter control flow divergence in compiled query pipelines“. Zusammen bilden sie einen einzigartigen Fingerprint.

Dieses zitieren