Parallel programming for OSEM reconstruction with MPI, OpenMP, and hybrid MPI-OpenMP

Jones, Mel; Yao, Rutao

doi:10.1109/nssmic.2004.1466323

Cited by 10 publications

(6 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Along the same line, Jones and Yau proposed a hybrid MPI and OpenMP approach to improve the parallel efficiency of the ordered‐subsets expectation‐maximization algorithm for reconstructing 3D positron emission tomography images. The algorithm reduced the interprocessor data exchange time, which was the dominant limiting factor of parallel efficiency of the MPI model when a large number of processors were used for the shared memory and the single system image distributed shared memory architectures.…”

Section: Related Workmentioning

confidence: 99%

An evaluation of MPI and OpenMP paradigms in finite‐difference explicit methods for PDEs on shared‐memory multi‐ and manycore systems

Cabral

Oliveira²,

Osthoff

et al. 2019

Concurrency and Computation

View full text Add to dashboard Cite

Summary This paper focuses on parallel implementations of three two‐dimensional explicit numerical methods on Intel® Xeon® Scalable Processor and the coprocessor Knights Landing. In this study, the performance of a hybrid parallel programming with message passing interface (MPI) and Open Multi‐Processing (OpenMP) and a pure MPI implementation used with two thread binding policies is compared with an improved OpenMP‐based implementation in three explicit finite‐difference methods for solving partial differential equations on shared‐memory multicore and manycore systems. Specifically, the improved OpenMP‐based version is a strategy that synchronizes adjacent threads and eliminates the implicit barriers of a naïve OpenMP‐based implementation. The experiments show that the most suitable approach depends on several characteristics related to the nonuniform memory access (NUMA) effect and load balancing, such as the size of the MPI domain and the number of synchronization points used in the parallel implementation. In algorithms that use four and five synchronization points, hybrid MPI/OpenMP approaches yielded better speedups than the other versions did in runs performed on both systems. The pure MPI‐based strategy, however, achieved better results than the other proposed approaches did in the method that employs only one synchronization point.

show abstract

Section: Related Workmentioning

confidence: 99%

An evaluation of MPI and OpenMP paradigms in finite‐difference explicit methods for PDEs on shared‐memory multi‐ and manycore systems

Cabral

Oliveira²,

Osthoff

et al. 2019

Concurrency and Computation

View full text Add to dashboard Cite

show abstract

“…However, such an approach reduces portability of codes. In a parallelization of a code for analysis of Positron Emission Tomography images [19] the authors conclude that a hybrid MPI-OpenMP approach yields optimal performance.…”

Section: Related Workmentioning

confidence: 99%

Exploiting Fine-Grained Parallelism in the Phylogenetic Likelihood Function with MPI, Pthreads, and OpenMP: A Performance Study

Stamatakis

Ott

2008

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Emerging multi-and many-core computer architectures pose new challenges with respect to efficient exploitation of parallelism. In addition, it is currently not clear which might be the most appropriate parallel programming paradigm to exploit such architectures, both from the efficiency as well as software engineering point of view. Beyond that, the application of high performance computing techniques and the use of supercomputers will be essential to deal with the explosive accumulation of sequence data. We address these issues via a thorough performance study by example of RAxML, which is a widely used Bioinformatics application for large-scale phylogenetic inference under the Maximum Likelihood criterion. We provide an overview over the respective parallelization strategies with MPI, Pthreads, and OpenMP and assess performance for these approaches on a large variety of parallel architectures. Results indicate that there is no universally best-suited paradigm with respect to efficiency and portability of the ML function. Therefore, we suggest that the ML function should be parallelized with MPI and Pthreads based on software engineering criteria as well as to enforce data locality.

show abstract

“…The projection based distribution using a replicated local image copy turned out to be the most promising approach for a parallelization of STIR as well. The works of [11] and [12] partially describe a parallel realization of the OSEM algorithm using MPI and OpenMP. Their implementation was only tested on a cluster system with dual-core nodes.…”

Section: A Related Workmentioning

confidence: 99%

A method for OSEM PET reconstruction on parallel architectures using STIR

Beisel

Lietsch

Thielemans³

2008

2008 IEEE Nuclear Science Symposium Conference Record

View full text Add to dashboard Cite

To accelerate image reconstruction of positron emission tomography (PET) data, we introduced an approach for parallel architectures by applying the message passing paradigm to an existing implementation of the ordered-subsets expectationmaximization (OSEM) algorithm for two-or three-dimensional (2D/3D) PET. To reduce the amount of time needed to complete a reconstruction, a cluster was used as well as different multicore systems. A main focus was on the multi-core processors, as these systems are increasingly common and easy to use in medical environments. The Open Source implementation 'Software for Tomographic Image Reconstruction' (STIR) was used as underlying reconstruction software, as it provides a wide range of usable scanner geometries and different algorithms. In particular STIR also offers an enhanced implementation of the OSEM algorithm. To allow for the use of arbitrary parallel systems, the standard message passing interface (MPI) was used to implement the parallel OSEM algorithm. Additionally, a special work package allocation algorithm was designed, that incorporates load balancing and improves the utilization of cached data. The resulting runtimes on up to 20 nodes and up to 8 cores were compared to the sequential runtime to determine the speedup and the parallel efficiency. The results using a limited set of test data achieved speedups of up to 8x, depending on the amount of data, the algorithm and the underlying architecture. We expect the algorithms to perform better for larger and more realistic data sets since the parallel overhead for distributing raw data and collecting the sub-results decreases in opposition to the actual computing time. The different parallel architectures were compared to determine the optimal system for PET Reconstruction. The cluster system achieved the best speedups using the available test data.

show abstract

Parallel programming for OSEM reconstruction with MPI, OpenMP, and hybrid MPI-OpenMP

Cited by 10 publications

References 5 publications

An evaluation of MPI and OpenMP paradigms in finite‐difference explicit methods for PDEs on shared‐memory multi‐ and manycore systems

An evaluation of MPI and OpenMP paradigms in finite‐difference explicit methods for PDEs on shared‐memory multi‐ and manycore systems

Exploiting Fine-Grained Parallelism in the Phylogenetic Likelihood Function with MPI, Pthreads, and OpenMP: A Performance Study

A method for OSEM PET reconstruction on parallel architectures using STIR

Contact Info

Product

Resources

About