Open MPI: Goals, Concept, and Design of a Next Generation MPI Implementation

Gabriel, Edgar; Fagg, Graham E.; Bosilca, George; Angskun, Thara; Dongarra, Jack; Squyres, Jeffrey M.; Sahay, Vishal; Kambadur, Prabhanjan; Barrett, Brian; Lumsdaine, Andrew; Castain, Ralph H.; Daniel, David; Graham, Richard L.; Woodall, Timothy S.

doi:10.1007/978-3-540-30218-6_19

Cited by 1,095 publications

(675 citation statements)

References 9 publications

Supporting

Mentioning

662

Contrasting

Unclassified

Order By: Relevance

“…We will be working first to prototype the MPI library side of the interface in both Open MPI [4] and MPICH 2 [5]. At the same time the parallel debugger side of the interface will be prototyped in the TotalView parallel debugger.…”

Section: Resultsmentioning

confidence: 99%

An Interface to Support the Identification of Dynamic MPI 2 Processes for Scalable Parallel Debugging

Gottbrath

Barrett²,

Gropp

et al. 2006

Recent Advances in Parallel Virtual Machine and Message Passing Interface

Self Cite

View full text Add to dashboard Cite

Abstract. This paper proposes an interface that will allow MPI 2 dynamic programs -those using MPI SPAWN, CONNECT/ACCEPT, or JOIN -to provide information to parallel debuggers such as TotalView about the set of processes that constitute an individual application. The TotalView parallel debugger currently obtains information about the identity of processes directly from the MPI library using a widely accepted proctable interface. The existing interface does not support MPI 2 dynamic operations. The proposed interface supports MPI 2 dynamic operations, subset debugging, and helps the parallel debugger assign meaningful names to processes.

show abstract

Section: Resultsmentioning

confidence: 99%

An Interface to Support the Identification of Dynamic MPI 2 Processes for Scalable Parallel Debugging

Gottbrath

Barrett²,

Gropp

et al. 2006

Recent Advances in Parallel Virtual Machine and Message Passing Interface

Self Cite

View full text Add to dashboard Cite

show abstract

“…Therefore, although we cannot leverage special hardware features, the protoype library is portable to all MPI-1 capable parallel computers. Further because we implemented optimized algorithms for all collective operations, we deliver the same performance as the hardware independent blocking collective operations in MPICH2 [10] and Open MPI 1.0 [11].…”

Section: Implementing Non-blocking Collective Operationsmentioning

confidence: 99%

Optimizing a conjugate gradient solver with non-blocking collective operations

et al. 2007

Self Cite

View full text Add to dashboard Cite

Abstract. This paper presents a case study about the applicability and usage of non-blocking collective operations. These operations provide the ability to overlap communication with computation and to avoid unnecessary synchronization. We introduce our NBC library, a portable lowoverhead implementation of non-blocking collectives on top of MPI-1. We demonstrate the easy usage of the NBC library with the optimization of a conjugate gradient solver with only minor changes to the traditional parallel implementation of the program. The optimized solver runs up to 34% faster and is able to overlap most of the communication. We show that there is, due to the overlap, no performance difference between Gigabit Ethernet and InfiniBand TM for our calculation.

show abstract

“…One difficulty with simple direct parameter measurement is the fact that most modern communication systems use message-size dependent protocols to optimize communication (e.g., [19,20]). Small messages are often copied to prepared (or pre-registered) local send or remote receive buffers to speed up the communication.…”

Section: Introductionmentioning

confidence: 99%

LogGP in theory and practice – An in-depth analysis of modern interconnection networks and benchmarking methods for collective operations

Hoefler

Schneider

Lumsdaine

2009

Simulation Modelling Practice and Theory

View full text Add to dashboard Cite

Accurate measurement and modeling of network performance is important for predicting and optimizing the running time of high-performance computing applications. Although the LogP family of models has proven to be a valuable tool for assessing the communication performance of parallel architectures, non-intrusive LogP parameter assessment of real systems remains a difficult task. Based on an analysis of accuracy and contention properties of existing measurement methods, we develop a new low-overhead measurement method which also assesses protocol changes in the underlying transport layers. We use the gathered parameters to simulate LogGP models of collective operations and demonstrate the errors in common benchmarking methods for collective operations. The simulations provide new insight into the nature of collective algorithms and their pipelining properties. We show that the error grows linearly with the system size.

show abstract

Open MPI: Goals, Concept, and Design of a Next Generation MPI Implementation

Cited by 1,095 publications

References 9 publications

An Interface to Support the Identification of Dynamic MPI 2 Processes for Scalable Parallel Debugging

An Interface to Support the Identification of Dynamic MPI 2 Processes for Scalable Parallel Debugging

Optimizing a conjugate gradient solver with non-blocking collective operations

LogGP in theory and practice – An in-depth analysis of modern interconnection networks and benchmarking methods for collective operations

Contact Info

Product

Resources

About