DocumentCode :
2721028
Title :
A multi-platform linear algebra toolbox for finite element solvers on heterogeneous clusters
Author :
Heuveline, Vincent ; Subramanian, Chandramowli ; Lukarski, Dimitar ; Weiss, Jan-Philipp
Author_Institution :
Eng. Math. & Comput. Lab. (EMCL), Karlsruhe Inst. of Technol., Karlsruhe, Germany
fYear :
2010
fDate :
20-24 Sept. 2010
Firstpage :
1
Lastpage :
6
Abstract :
Heterogeneous clusters with multiple sockets and multicore-processors accelerated by dedicated coprocessors like GPUs, Cell BE, FPGAs or others nowadays provide unrivaled computing power in terms of floating point operations. Specific capabilities of additional processor technologies enable dedicated exploitation with respect to particular application and data characteristics. However, resource utilization, programmability, and scalability of applications across heterogeneous platforms is a major concern. In the framework of the HiFlow finite element software package we have developed a portable software approach that implements efficient parallel solvers for partial differential equations by means of unified and modular user interfaces across a variety of heterogeneous platforms - in particular on GPU accelerated clusters. We detail our concept and provide performance analysis for various test scenarios that prove performance capabilities, scalability, viability, and user friendliness.
Keywords :
computer graphic equipment; coprocessors; field programmable gate arrays; finite element analysis; linear algebra; mathematics computing; multiprocessing systems; partial differential equations; resource allocation; user interfaces; Cell BE; FPGA; GPU; HiFlow finite element software package; coprocessors; finite element solvers; floating point operations; heterogeneous clusters; multicore processors; multiplatform linear algebra toolbox; parallel solvers; partial differential equations; programmability; resource utilization; scalability; user interfaces; Multicore processing; Heterogeneous clusters; conjugate gradient; multi-GPU; performance analysis; scalability; sparse linear algebra; speedup;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Cluster Computing Workshops and Posters (CLUSTER WORKSHOPS), 2010 IEEE International Conference on
Conference_Location :
Heraklion, Crete
Print_ISBN :
978-1-4244-8395-2
Electronic_ISBN :
978-1-4244-8397-6
Type :
conf
DOI :
10.1109/CLUSTERWKSP.2010.5613084
Filename :
5613084
Link To Document :
بازگشت