A library for parallel thread-level speculation

Aguilar, José; Campero, Kahlil

doi:10.1109/clei.2013.6670596

Cited by 1 publication

(6 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For 183.equake, in our study obtains an execution on average 1.9 times faster than the original program. This value is lower than that obtained by Prabhu for a real memory system with 8 cores (~ 2.25x), but better with respect to results obtained in [15]. But, the percentage of memory used by smvp routine is reduced to 61.55% in our study.…”

Section: Experiments 5 Definition Of the Experimental Platformcontrasting

confidence: 84%

“…In this occasion we are obtained significant improvement in the use of memory with our speculative models. Particularly, we observe that has a favorable effect on the parallelization in the case of 183.equake, but not in the case of 179.art, because we have improved the previous results obtained in [15]. Figures 3 and 4 show the memory usage for the parallelization carried out in [15] and the new version of parallelization carried out during this study.…”

Section: Experiments 5 Definition Of the Experimental Platformmentioning

confidence: 51%

“…Particularly, we observe that has a favorable effect on the parallelization in the case of 183.equake, but not in the case of 179.art, because we have improved the previous results obtained in [15]. Figures 3 and 4 show the memory usage for the parallelization carried out in [15] and the new version of parallelization carried out during this study. [15] We have improved the memory usage in our new version of parallel program for the case of 183.equake, due to number of copies generated (we have reduced the number of threads to manage the critical sections and the "repeat for" instructions of the internal loops of the smvp routine).…”

Section: Experiments 5 Definition Of the Experimental Platformmentioning

confidence: 51%

“…The running time for 179.art was 174 seconds, which represents an acceleration of ~ 2.2x with respect to the base case. In this new version, we also have improved the memory usage with respect to [15], but we obtain the same accelerations. The acceleration obtained for Prabhu 179.art is second ~ 2.25x values, similar to our results because we have improved our parallel version of 179.art algorithm.…”

Section: Experiments 5 Definition Of the Experimental Platformmentioning

confidence: 87%

“…The library proposed in [14,15] offers speculative functionalities to a programmer. The library consists of two layers: an outer layer and an inner layer, which correspond to the programming logic and data management.…”

Section: A Library For Explicit Parallelism Based On Thread-level Spementioning

confidence: 99%

See 4 more Smart Citations

An explicit parallelism study based on thread-level speculation

Aguilar

Campero

2014

CLEIej

Self Cite

View full text Add to dashboard Cite

Developments in parallel architectures are an important branch in computer science. The success of such architectures derives from their inherent ability to improve the program performances. However, their ability to improve the performance on programs depends on the parallelism extraction strategies, which are always limited by the logic of each sequential program. Speculation is the only known alternative to overcome these constraints and increase the parallelism. In this paper, we study the explicit speculative parallelism using a library of thread-level speculation. We present the design of this library and study different speculative models: speculation of decision structures, speculation of loops, and speculation of critical sections. Our study evaluates different cases taken from SPEC CPU 2000, allowing acceleration of about 1.8x in multicore architectures (four core) with coarse-grained multithreaded.

show abstract

Section: Experiments 5 Definition Of the Experimental Platformcontrasting

confidence: 84%

Section: Experiments 5 Definition Of the Experimental Platformmentioning

confidence: 51%