The continuously increasing size of biological sequence databases has motivated the development of analysis suites that, by means of parallelization, are capable of performing faster searches on such databases. Howeve...
详细信息
The continuously increasing size of biological sequence databases has motivated the development of analysis suites that, by means of parallelization, are capable of performing faster searches on such databases. However, many of these tools are not suitable for execution on mid-to-large scale parallel infrastructures such as computational Grids. This paper shows how COMP Superscalar can be used to effectively parallelize on the Grid a sequence analysis program. In particular, we present a sequential version of the HMMER hmmpfam tool that, when run with COMP Superscalar, is decomposed into tasks and run on a set of distributed resources, not burdening the programmer with parallelization efforts. Although performance is not a main objective of this work, we also present some test results where COMP Superscalar, using a new pre-scheduling technique, clearly outperforms a well-known parallelization of the hmmpfam algorithm.
Big Data analysis refers to advanced and efficient datamining and machine learning techniques applied to large amount of data. Research work and results in the area of Big Data analysis are continuously rising, and mo...
详细信息
Big Data analysis refers to advanced and efficient datamining and machine learning techniques applied to large amount of data. Research work and results in the area of Big Data analysis are continuously rising, and more and more new and efficient architectures, programmingmodels, systems, and data mining algorithms are proposed. Taking into account the most popular programmingmodels for Big Data analysis (MapReduce, Directed Acyclic Graph, Message Passing, Bulk Synchronous parallel, Workflow and SQL-like), we analysed the features of the main systems implementing them. Such systems are compared using four classification criteria (i.e. level of abstraction, type of parallelism, infrastructure scale and classes of applications) for helping developers and users to identify and select the best solution according to their skills, hardware availability, productivity and application needs. [GRAPHICS] This figure is a word cloud highlighting the most popular words related to Big Data analysis.
Many-core heterogeneous designs are nowadays widely available among embedded systems. Initiatives such as the HSA push for a model where the host processor and the accelerator(s) communicate via coherent, Unified Virt...
详细信息
ISBN:
(纸本)9781450350396
Many-core heterogeneous designs are nowadays widely available among embedded systems. Initiatives such as the HSA push for a model where the host processor and the accelerator(s) communicate via coherent, Unified Virtual Memory (UVM). In this paper we describe our experience in porting the OpenMP v4 programming model to a low-end, heterogeneous embedded system based on the PULP many-core accelerator featuring lightweight (software-managed) UVM support. We describe a GCC-based toolchain which enables: i) the automatic generation of host and accelerator binaries from a single, high-level, OpenMP parallel program; ii) the automatic instrumentation of the accelerator program to transparently manage UVM. This enables up to 4x faster execution compared to traditional copy-based offload mechanisms.
暂无评论