Off-loading compute intensive tasks for insurance products using a just-in-time compiler on a hybrid system

Authors:
Thomas Grosser;Alexander C. Schmid;Markus Deuling;Hoang-Nam Nguyen;Wolfgang Rosenstiel
Affiliations:
University of Tübingen;University of Tübingen;University of Tübingen;University of Tübingen;University of Tübingen
Venue:
CASCON '09 Proceedings of the 2009 Conference of the Center for Advanced Studies on Collaborative Research
Year:
2009

Citing 3
Cited 0

LLVM: A Compilation Framework for Lifelong Program Analysis & Transformation

Proceedings of the international symposium on Code generation and optimization: feedback-directed and runtime optimization
High performance combinatorial algorithm design on the Cell Broadband Engine processor

Parallel Computing
Compilers: Principles, Techniques, & Tools with Gradiance

Compilers: Principles, Techniques, & Tools with Gradiance

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper focuses on building a hybrid system, which consists of IBM System z and IBM POWER processor-based blades as accelerators. The business application under investigation is an insurance software that is used to perform batch runs and thus requires valuable compute cycles on System z. Therefore, we present a novel technology to accelerate mathematical insurance calculations by implementing a just-in-time compiler based on LLVM. The resulting code is optimized for network attached accelerator platforms and thus runs faster and more efficiently. Overall, the hybrid solution is able to achieve a higher calculation throughput as opposed to a System z-only solution. We are able to achieve a high speed-up compared to interpretation with simple mathematical calculations.