Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userAuto Adaptive Strategy for Parallel Applications
استراتيجية ذاتية التكيف للتطبيقات المتوازية
1352
0 44
0
(
0
)
Created by
Shamra Editor
Ask ChatGPT about the research
سنقدم في هذا البحث استراتيجية ذاتية التكيف تمكننا من كتابة خوارزمية متوازية تتكيف مع عدد الموارد المتوفرة على البيئة التفرعية المخصصة لتنفيذ البرنامج المتوازي. إن التطبيقات المتوازية المدروسة و المعنية بالبحث هي تطبيقات ممثلة بمخطط تدفق البيانات المبني ديناميكياً خلال التنفيذ. تقوم الطريقة المقترحة هنا على المزاوجة بين خوارزمية تسلسلية و أخرى متوازية معتمدين على مبدأ سرقة العمل في جدولة المهام. و نقدم دراسة لتعقيد هذه الخوارزمية المتكيفة و تحليل لأدائها على معالج و مقارنته مع خوارزمية تفرعية تقليدية.
We introduce an auto adaptive strategy enables to write a parallel algorithm adapts to the number of available resources at allocated parallel environment to execute the parallel program. The parallel applications we are studying which are represented by data-flow graph which built dynamically during the execution. The new suggested strategy is based on coupling of a sequential algorithm and a parallel one and relies on the principle of work stealing in the tasks scheduling. We offer a study of the complexity of the adaptive algorithm and analyze its performance on processors and compare it with a performance of a classic parallel algorithm.
References used
BENDER MA, RABIN MO, 2002, Online scheduling of parallel programs on heterogeneous systems with applications to cilk, Theory Comput Syst. Vol 35.3, pp 289-304
BERNARD J, TRAORE D, ROCH JL, 2006, On-line adaptive parallel prefix computation, Euro-Par, vol 4128, pp 841-850
GALILEE F, CAVALHEIRO G, DOREILLE M, ROCH JL, 1998, Athapascan-1: On-Line Building Data Flow Graph in a Parallel Language, International Conference on Parallel Architectures and Compilation Techniques, PACT'98, pp 88–95
rate research
Read More
The work aims to make benefit from existence
multi-CPU and multi-GPU, exploiting the calculation processes
which do multi-GPU, which aims to form mechanism to
scheduling a directed acyclic graph(DAG), it aims to reduce
communication between resources and inter linked task
scheduling in the best form.
In this paper, we introduce a continuous mathematical model to
optimize the compromise between the overhead of fault tolerance
mechanism and the faults impacts in the environment of
execution. The fault tolerance mechanism considered in this
rese
arch is a coordinated checkpoint/recovery mechanism and the
study based on stochastic model of different performance critics of
parallel application on parallel and distributed environment.
In this research, We introduce two probabilistic mechanisms to
certificate parallel applications on distribute architecture supposing
that there are no oracles on which we depend on certification, in
addition to introducing cost model of two mecha
nisms and compare
them.
In this research, we are interested in parallel applications, which
are represented by data-flow graph that is built dynamically during
the execution and which are executed in a wide distributed
heterogeneous and dynamic environment and these applications
use the principle of work stealing to distribute the tasks among the
processors.
In this paper we present a study on the time cost
added to the grid computing as a result of the use of a
coordinated checkpoint / recovery fault tolerance protocol, we aim
to find a mathematical model which determined the suitable time
to save t
he checkpoints for application, to achieve a minimum
finish time of parallel application in grid computing with faults and
fault tolerance protocols, we have find this model by serial
modeling to the goal errors, execution environment and the
chosen fault tolerance protocol all that by Kolmogorov differential
equations.
We propose a novel framework to train models to classify acceptability of responses generated by natural language generation (NLG) models, improving upon existing sentence transformation and model-based approaches. An NLG response is considered accep
table if it is both semantically correct and grammatical. We don't make use of any human references making the classifiers suitable for runtime deployment. Training data for the classifiers is obtained using a 2-stage approach of first generating synthetic data using a combination of existing and new model-based approaches followed by a novel validation framework to filter and sort the synthetic data into acceptable and unacceptable classes. Our 2-stage approach adapts to a wide range of data representations and does not require additional data beyond what the NLG models are trained on. It is also independent of the underlying NLG model architecture, and is able to generate more realistic samples close to the distribution of the NLG model-generated responses. We present results on 5 datasets (WebNLG, Cleaned E2E, ViGGO, Alarm, and Weather) with varying data representations. We compare our framework with existing techniques that involve synthetic data generation using simple sentence transformations and/or model-based techniques, and show that building acceptability classifiers using data that resembles the generation model outputs followed by a validation framework outperforms the existing techniques, achieving state-of-the-art results. We also show that our techniques can be used in few-shot settings using self-training.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
