Evaluation of OpenMP for the Cyclops Multithreaded Architecture

Abstract

Multithreaded architectures have the potential of tolerating large memory and functional unit latencies and increase resource utilization. The Blue Gene/Cyclops architecture, being developed at the IBM T. J. Watson Research Center, is one such systems that offers massive intra-chip parallelism. Although the BG/C architecture was initially designed to execute specific applications, we believe that it can be effectively used on a broad range of parallel numerical applications. Programming such applications for this unconventional design requires a significant porting effort when using the basic built-in mechanisms for thread management and synchronization. In this paper, we describe the implementation of an OpenMP environment for parallelizing applications, currently under development at the CEPBA-IBM Research Institute, targeting BG/C. The environment is evaluated with a set of simple numerical kernels and a subset of the NAS OpenMP benchmarks. We identify issues that were not initially considered in the design of the BG/C architecture to support a programming model such as OpenMP. We also evaluate features currently offered by the BG/C architecture that should be considered in the implementation of an efficient OpenMP layer for massive intra-chip parallel architectures.

An Evaluation of OpenMP on Current and Emerging Multithreaded/Multicore Processors

“…The reported values are for 2 threads per processor and have been normalized with respect to the single-thread per processor execution of each benchmark on the specific architecture and number of processors. This way, the graphs emphasize the effects of using a second 1 All the NAS applications we used are iterative. The computational routines are enclosed in an external, sequential loop.…”

“…Earlier research efforts have ported and evaluated OpenMP on specific processor designs, including heterogeneous chip multiprocessors [14], slipstream processors [9] (a form of 2-way chip multiprocessors in which the second core is used for speculative runahead execution) and Cyclops, a fine-grain multithreaded processor architecture introduced by IBM [1]. Our evaluation focuses on commodity processors, with organizations spanning the design space between simultaneous multithreading and chip multiprocessors and a few execution contexts.…”

“…Full system simulation with Simics introduces an average 7000-fold slowdown in the execution time of applications, compared with the execution on a real machine. We simulated the same application binaries, using the same data sets, however we reduced the number of iterations 1 we ran on the simulator in order to limit the execution time to reasonable levels. More specifically, we executed only 3 of the outermost iterations of each benchmark, discarding the results from the first iteration in order to eliminate transient effects due to cache warmup.…”

Optimizing NANOS OpenMP for the IBM Cyclops Multithreaded Architecture

“…In [1] the authors evaluated the performance of some of the NAS benchmarks on Cyclops. The performance metric used to evaluate performance is speedup with respect to the sequential execution.…”

“…Our tools rely on the Cyclops developer toolkit and simulator to evaluate performance. An initial evaluation of the porting was presented in [1]. This paper proposes two different kinds of optimizations (software and hardware) to run applications on the Cyclops architecture.…”