Memory Bandwidth: The True Bottleneck of SIMD Multimedia Performance on a Superscalar Processor

Authors:
Julien Sébot;Nathalie Drach-Temam
Affiliations:
-;-
Venue:
Euro-Par '01 Proceedings of the 7th International Euro-Par Conference Manchester on Parallel Processing
Year:
2001

Citing 4
Cited 0

Exploiting instruction level parallelism in geometry processing for three dimensional graphics applications

MICRO 31 Proceedings of the 31st annual ACM/IEEE international symposium on Microarchitecture
Evaluating MMX technology using DSP and multimedia applications

MICRO 31 Proceedings of the 31st annual ACM/IEEE international symposium on Microarchitecture
Performance of image and video processing with general-purpose processors and media ISA extensions

ISCA '99 Proceedings of the 26th annual international symposium on Computer architecture
Exploiting SIMD parallelism in DSP and multimedia algorithms using the AltiVec technology

ICS '99 Proceedings of the 13th international conference on Supercomputing

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents the performance of DSP, image and 3D applications on recent general-purpose microprocessors using streaming SIMD ISA extensions (integer and floating point). The 9 benchmarks benchmark we use for this evaluation have been optimized for DLP and caches use with SIMD extensions and data prefetch. The result of these cumulated optimizations is a speedup that ranges from 1.9 to 7.1. All the benchmarks were originaly computation bound and 7 becomes memory bandwidth bound with the addition of SIMD and data prefetch. Quadrupling the memory bandwidth has no effect on original kernels but improves the performance of SIMD kernels by 15-55%.