Challenges of Scaling Algebraic Multigrid Across Modern Multicore Architectures

Baker, Allison H.; Gamblin, Todd; Schulz, Martin; Yang, Ulrike Meier

doi:10.1109/ipdps.2011.35

Cited by 56 publications

(57 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We present total times in Figure 4. Separate setup and cycle times for this problem on the three architectures are described in the original paper [3].…”

Section: Scaling Of Amgmentioning

confidence: 99%

See 1 more Smart Citation

Scaling Algebraic Multigrid Solvers: On the Road to Exascale

Baker

Falgout

Gamblin

et al. 2011

Competence in High Performance Computing 2010

Self Cite

View full text Add to dashboard Cite

Algebraic Multigrid (AMG) solvers are an essential component of many large-scale scientific simulation codes. Their continued numerical scalability and efficient implementation is critical for preparing these codes for exascale. Our experiences on modern multi-core machines show that significant challenges must be addressed for AMG to perform well on such machines. We discuss our experiences and describe the techniques we have used to overcome scalability challenges for AMG on hybrid architectures in preparation for exascale.

show abstract

“…We present total times in Figure 4. Separate setup and cycle times for this problem on the three architectures are described in the original paper [3].…”

Section: Scaling Of Amgmentioning

confidence: 99%

“…This effect is still visible, but less pronounced in the solve phase, which has a smaller amount of communication. On Hera, the worst performance for the solve phase is obtained for H1x16 (see [3,4]), and is caused by the NUMA architecture. Using the MCSup library, see H1x16MC, performance is significantly improved.…”

Section: Scaling Of Amgmentioning

confidence: 99%

Scaling Algebraic Multigrid Solvers: On the Road to Exascale

Baker

Falgout

Gamblin

et al. 2011

Competence in High Performance Computing 2010

Self Cite

View full text Add to dashboard Cite

show abstract

“…The Mcsup library [2] has been used to perform NUMA aware data allocations. It works by trapping all allocations of memory and making sure data is in the same NUMA domain as the task that accesses it.…”

Section: Numa Aware Data Allocationmentioning

confidence: 99%

LULESH Programming Model and Performance Ports Overview

Karlin¹

2012

View full text Add to dashboard Cite

“…Figure 1 illustrates the challenge. Under the current state-of-practice, a user running GTC [37], a plasma physics code, pays 60% more when co-located with LAMMPS [34], a molecular dynamics code, versus AMG [4,6], a parallel algebraic multigrid solver. To remedy this problem, we suggest discounting a user based on the interference caused by the other corunning applications.…”

Section: Introductionmentioning

confidence: 99%

Enabling Fair Pricing on High Performance Computer Systems with Node Sharing

Breslow

Tiwari

Schulz

et al. 2014

Scientific Programming

Self Cite

View full text Add to dashboard Cite

Abstract. Co-location, where multiple jobs share compute nodes in large-scale HPC systems, has been shown to increase aggregate throughput and energy efficiency by 10-20%. However, system operators disallow co-location due to fair-pricing concerns, i.e., a pricing mechanism that considers performance interference from co-running jobs. In the current pricing model, application execution time determines the price, which results in unfair prices paid by the minority of users whose jobs suffer from co-location. This paper presents POPPA, a runtime system that enables fair pricing by delivering precise online interference detection and facilitates the adoption of supercomputers with co-locations. POPPA leverages a novel shutter mechanism -a cyclic, fine-grained interference sampling mechanism to accurately deduce the interference between co-runners -to provide unbiased pricing of jobs that share nodes. POPPA is able to quantify inter-application interference within 4% mean absolute error on a variety of co-located benchmark and real scientific workloads.

show abstract

Challenges of Scaling Algebraic Multigrid Across Modern Multicore Architectures

Cited by 56 publications

References 9 publications

Scaling Algebraic Multigrid Solvers: On the Road to Exascale

Scaling Algebraic Multigrid Solvers: On the Road to Exascale

LULESH Programming Model and Performance Ports Overview

Enabling Fair Pricing on High Performance Computer Systems with Node Sharing

Contact Info

Product

Resources

About