Composable, non-blocking collective operations on power7 IH

Tanase, Gabriel; Almasi, Gheorghe; Xue, Hanhong; Archer, Charles J

doi:10.1145/2304576.2304605

Cited by 10 publications

(10 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The new code transformations are prototyped in the XLUPC compiler framework [32] that uses the IBM PGAS runtime [33]. The compiler contains additional optimizations for UPC and other languages, including C and C++.…”

Section: Methodsmentioning

confidence: 99%

Combining Static and Dynamic Data Coalescing in Unified Parallel C

Alvanos¹,

Farreras

Tiotto

et al. 2016

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

Abstract-Significant progress has been made in the development of programming languages and tools that are suitable for hybrid computer architectures that group several shared-memory multicores interconnected through a network. This paper addresses important limitations in the code generation for Partitioned Global Address Space (PGAS) languages. These languages allow fine-grained communication and lead to programs that perform many fine-grained accesses to data. When the data is distributed to remote computing nodes, code transformations are required to prevent performance degradation. Until now code transformations to PGAS programs have been restricted to the cases where both the physical mapping of the data or the number of processing nodes are known at compilation time. In this paper, a novel application of the inspector-executor model overcomes these limitations and allows profitable code transformations, which result in fewer and larger messages sent through the network, when neither the data mapping nor the number of processing nodes are known at compilation time. A performance evaluation reports both scaling and absolute performance numbers on up to 32768 cores of a Power 775 supercomputer. This evaluation indicates that the compiler transformation results in speedups between 1.15X and 21X over a baseline and that these automated transformations achieve up to 63% the performance of the MPI versions.

show abstract

Section: Methodsmentioning

confidence: 99%

Combining Static and Dynamic Data Coalescing in Unified Parallel C

Alvanos¹,

Farreras

Tiotto

et al. 2016

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

show abstract

“…PAMI collective library builds upon the portable Component Collective Messaging Interface (CCMI [4]) from Blue Gene/P, with optimized implementations on Blue Gene/Q, Power7 IH [6] and Intel x86 clusters.…”

Section: Pami Collectivesmentioning

confidence: 99%

Optimization of MPI_Allreduce on the blue Gene/Q supercomputer

Kumar

Faraj

2013

Proceedings of the 20th European MPI Users' Group Meeting

View full text Add to dashboard Cite

The IBM Blue Gene/Q supercomputer has a 5D torus network where each node is connected to ten bi-directional links. In this paper we present techniques to optimize the MPI Allreduce collective operation by building ten different edge disjoint spanning trees on the ten torus links. We accelerate summing of network packets with local buffers by the use of Quad Processing SIMD unit in the BG/Q cores and executing the sums on multiple communication threads created by the PAMI libraries. The net gain we achieve is a peak throughput of 6.3 GB/sec for double precision floating point sum allreduce, that is a speedup of 3.75x over the collective network based algorithm in the product MPI stack on BG/Q.

show abstract

“…Optimizations are possible for both point to point messages and collectives. Very good performance can often be attained using this approach [18]. This, of course, still leaves the system with a very large number of processes running, which can cause bottlenecks.…”

Section: Related Workmentioning

confidence: 99%

“…A major difference is the completion notification using call backs in PAMI versus events in CCI. In a related paper [18] we discuss the importance of callbacks for non blocking primitives (P2P, collective) composition. Another important difference relative to the results reported in [6] is that this paper focuses especially on a multi-threaded deployment while the CCI paper only present result for the distributed memory case.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Network Endpoints for Clusters of SMPs

Tanase

Almasi

Xue³

et al. 2012

2012 IEEE 24th International Symposium on Computer Architecture and High Performance Computing

Self Cite

View full text Add to dashboard Cite

Modern large scale parallel machines feature an increasingly deep hierarchy of interconnections. Individual processing cores employ simultaneous multithreading (SMT) to better exploit functional units; multiple coherent processors are collocated in a node to better exploit links to cache, memory and network (SMP); and multiple nodes are interconnected by specialized low latency/high speed networks. Current trends indicate ever wider SMP nodes in the future. To service these nodes, modern high performance network devices (including Infiniband and all of IBM's recent offerings) offer the ability to sub-divide the network devices' resources among the processing threads. System software, however, lags in exploiting these capabilities, leaving users of e.g., MPI[14], UPC[19] in a bind, requiring complex and fragile workarounds in user programs.In this paper we discuss our implementation of endpoints, the software paradigm central to the IBM PAMI messaging library [3]. A PAMI endpoint is an expression in software of a slice of the network device. System software can service endpoints without serializing the many threads on an SMP by forcing them through a critical section. In the paper we describe the basic guarantees offered by PAMI to the programmer, and how these can be used to enable efficient implementations of high level libraries and programming languages like UPC. We evaluate the efficiency of our implementation on a novel P7IH system with up to 4096 cores, running microbenchmarks designed to find performance deficiencies in the endpoints implementation of both point-to-point and collective functions.

show abstract

Composable, non-blocking collective operations on power7 IH

Cited by 10 publications

References 19 publications

Combining Static and Dynamic Data Coalescing in Unified Parallel C

Combining Static and Dynamic Data Coalescing in Unified Parallel C

Optimization of MPI_Allreduce on the blue Gene/Q supercomputer

Network Endpoints for Clusters of SMPs

Contact Info

Product

Resources

About