Write a Blog >>
SPLASH 2017
Sun 22 - Fri 27 October 2017 Vancouver, Canada
Wed 25 Oct 2017 10:52 - 11:15 at Regency C - Performance Chair(s): Kathryn S McKinley

Heterogeneous architectures characterize today hardware ranging from super-computers to smartphones. However, in spite of this importance, programming such systems is still challenging. In particular, it is challenging to map computations to the different processors of a heterogeneous device. In this paper, we provide a static analysis that mitigates this problem. Our contributions are two-fold: first, we provide a semi-context-sensitive algorithm, which analyzes the program's call graph to determine the best processor for each calling context. This algorithm is parameterized by a cost model, which takes into consideration processor's characteristics and data transfer time. Second, we show how to use simulated annealing to calibrate this cost model for a given heterogeneous architecture. We have used our ideas to build Etino, a tool that annotates C programs with OpenACC or OpenMP 4.0 directives. Etino generates code for a CPU-GPU architecture without user intervention. Experiments on classic benchmarks reveal speedups of up to 75x. Moreover, our calibration process lets \etino{} avoid slowdowns of up to 720x which trivial parallelization approaches would yield.

Wed 25 Oct

Displayed time zone: Tijuana, Baja California change

10:30 - 12:00
PerformanceOOPSLA at Regency C
Chair(s): Kathryn S McKinley Google
10:30
22m
Talk
A Volatile-by-Default JVM for Server Applications
OOPSLA
Lun Liu University of California at Los Angeles, USA, Todd Millstein University of California, Los Angeles, Madan Musuvathi Microsoft Research
DOI
10:52
22m
Talk
Static Placement of Computation on Heterogeneous Devices
OOPSLA
Gabriel Poesia Federal University of Minas Gerais, Brazil, Breno Campos Ferreira Guimarães Federal University of Minas Gerais, Brazil, Fabrício Ferracioli LG Electronics, Brazil, Fernando Magno Quintão Pereira UFMG
DOI
11:15
22m
Talk
Skip Blocks: Reusing Execution History to Accelerate Web Scripts
OOPSLA
Sarah E. Chasins University of California, Berkeley, Rastislav Bodík University of Washington
DOI
11:37
22m
Talk
Virtual Machine Warmup Blows Hot and Cold
OOPSLA
Edd Barrett King's College London, CF Bolz-Tereick , Rebecca Killick Department of Mathematics and Statistics, University of Lancaster, Sarah Mount King's College London, Laurence Tratt King's College London
DOI