1 / 17

PBB: A Parallel Bioinformatics Benchmark Suite for Shared Memory Multiprocessors

PBB: A Parallel Bioinformatics Benchmark Suite for Shared Memory Multiprocessors. CHEN Wenguang HPC Inst., CS Dept., Tsinghua University. Outlines. Motivation Benchmark selection & construction Benchmark characteristics Performance results Conclusions, Q&A. Motivation.

salvatorem
Télécharger la présentation

PBB: A Parallel Bioinformatics Benchmark Suite for Shared Memory Multiprocessors

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. PBB: A Parallel Bioinformatics Benchmark Suite for Shared Memory Multiprocessors CHEN Wenguang HPC Inst., CS Dept., Tsinghua University

  2. Outlines • Motivation • Benchmark selection & construction • Benchmark characteristics • Performance results • Conclusions, Q&A

  3. Motivation • Widely use of bioinformatics applications • The trend of multi-core => There should be a parallel bioinfo. benchmark • SPEC CPU2000 may not match the characteristics of bioinformatics workloads well • Existing bioinformatics benchmark are not satisfactory => We need a new one

  4. Existing Bioinformatics Benchmarks: BioBench: • does not cover some important domains • no parallel program BioPerf: • includes only one parallel benchmark BioParallel: our previous work, includes 5 parallel application

  5. PBB benchmark suite: Being more complete 7 applications covering 7 of the most important domains of bioinfo. Keeping pace with the changing world all the applications are parallelized

  6. Benchmark Selection & Construction • Identify the most important application domains • Choose representative applications for each domain most popular, most advanced • Benchmark optimization & parallelization

  7. The 7 applications: • Pairwise sequence alignment: BLAST-P • Global alignment: PLSA • Multiple sequences alignment: MUSCLE • Protein 3D structure prediction: Rosetta • Phylogenetic tree reconstruction: SEMPHY • Gene regulatory network learning: ModuleNet • Pattern study of Single Nucleotide Polymorphisms: SNP

  8. Benchmark Characteristics Systems Used: Workload analysis is performed on QP001

  9. Non-egligible FP Instruction profile: Higher L/S

  10. CPI Low CPI 8P means QP001 with HT enabled

  11. FSB bandwidth utilization Low utilization

  12. Performance Results Benchmark scores: : time used to run application i on the reference system : time used to run application i on the tested system Rosetta is excluded for it produces random results

  13. Scores of PBB

  14. Parallel speedup Tested on Unisys-ES700 with 16 Xeon

  15. Hyper Threading Effects

  16. Conclusions Workload characteristics: • High percentage of load/store instructions • Non-negligible floating point instructions, but still significantly lower than SPECCPU 2000FP • Low CPI • Low memory bandwidth demand

  17. Thanks Any questions?

More Related