FG-MPI: Fine-grain MPI for multicore and clusters

Kamal, Humaira; Wagner, Alan

doi:10.1109/ipdpsw.2010.5470773

Cited by 32 publications

(18 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…More recent work in FG-MPI [9] shares the same idea with AMPI by exploiting fine grained decomposition using threads. However, FG-MPI does not support thread migration and dynamic load balancing.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Automatic MPI to AMPI Program Transformation Using Photran

Negara

Zheng

Pan

et al. 2011

Euro-Par 2010 Parallel Processing Workshops

View full text Add to dashboard Cite

Abstract. Adaptive MPI, or AMPI, is an implementation of the Message Passing Interface (MPI) standard. AMPI benefits MPI applications with features such as dynamic load balancing, virtualization, and checkpointing. Because AMPI uses multiple user-level threads per physical core, global variables become an obstacle. It is thus necessary to convert MPI programs to AMPI by eliminating global variables. Manually removing the global variables in the program is tedious and error-prone. In this paper, we present a Photran-based tool that automates this task with a source-to-source transformation that supports Fortran. We evaluate our tool on the multi-zone NAS Benchmarks with AMPI. We also demonstrate the tool on a real-world large-scale FLASH code and present preliminary results of running FLASH on AMPI. Both results show significant performance improvement using AMPI. This demonstrates that the tool makes using AMPI easier and more productive.

show abstract

Section: Related Workmentioning

confidence: 99%

“…Dynamic load balancing is supported via thread migration. More recent work in FG-MPI [9] also follows this direction; however, it does not support thread migration and dynamic load balancing yet.…”

Section: Introductionmentioning

confidence: 99%

Automatic MPI to AMPI Program Transformation Using Photran

Negara

Zheng

Pan

et al. 2011

Euro-Par 2010 Parallel Processing Workshops

View full text Add to dashboard Cite

show abstract

“…Adaptive MPI (AMPI) [4], [5], FG-MPI [6], Phoenix [7] and TMPI (Threaded MPI) [8] exemplify this approach. One advantage of this approach is that it allows automatic adaptive overlap of communication and computationwhen one MPI thread is blocked to receive a message, another MPI thread on the same processor can be scheduled to be executed.…”

Section: Motivationmentioning

confidence: 99%

“…More recent work, in FG-MPI [6], shares the same idea with AMPI by exploiting fine-grained decomposition using threads. However, FG-MPI does not support thread migration and dynamic load balancing.…”

Section: Flashmentioning

confidence: 99%

Automatic Handling of Global Variables for Multi-threaded MPI Programs

Zheng

Negara

Mendes

et al. 2011

2011 IEEE 17th International Conference on Parallel and Distributed Systems

View full text Add to dashboard Cite

Abstract-Hybrid programming models, such as MPI combined with threads, are one of the most efficient ways to write parallel applications for current machines comprising multi-socket/multicore nodes and an interconnection network. Global variables in legacy MPI applications, however, present a challenge because they may be accessed by multiple MPI threads simultaneously. Thus, transforming legacy MPI applications to be thread-safe in order to exploit multi-core architectures requires proper handling of global variables.In this paper, we present three approaches to eliminate global variables to ensure thread-safety for an MPI program. These approaches include: (a) a compiler-based refactoring technique, using a Photran-based tool as an example, which automates the source-to-source transformation for programs written in Fortran; (b) a technique based on a global offset table (GOT); and (c) a technique based on thread local storage (TLS). The second and third methods automatically detect global variables and privatize them for each thread at runtime. We discuss the advantages and disadvantages of these approaches and compare their performance using both synthetic benchmarks, such as the NAS Benchmarks, and a real scientific application, the FLASH code.

show abstract

“…We introduced Fine-Grain MPI (FG-MPI) to investigate the extent to which function-level parallelism can be supported while staying within the MPI process model [8,9]. To this end, FG-MPI augments the MPI middleware inside each OS-process with a runtime that supports hundreds and thousands of finergrain MPI processes inside an OS-process.…”

Section: Introductionmentioning

confidence: 99%

An Integrated Runtime Scheduler for MPI

Kamal

Wagner

2012

Recent Advances in the Message Passing Interface

Self Cite

View full text Add to dashboard Cite

Abstract. Fine-Grain MPI (FG-MPI) supports function-level parallelism while staying within the MPI process model. It provides a runtime that is directly integrated into the MPICH2 middleware and uses lightweight coroutines to implement an MPI-aware scheduler. Our key observation is that having multiple MPI processes per OS-process, with a runtime scheduler can be used to simplify MPI programming and achieve performance without adding complexity to the program. The performance part of the program is now outside of the specification of the program in the runtime where performance can be tuned with few, if any, changes to the code.

show abstract

FG-MPI: Fine-grain MPI for multicore and clusters

Cited by 32 publications

References 14 publications

Automatic MPI to AMPI Program Transformation Using Photran

Automatic MPI to AMPI Program Transformation Using Photran

Automatic Handling of Global Variables for Multi-threaded MPI Programs

An Integrated Runtime Scheduler for MPI

Contact Info

Product

Resources

About