| Attributes | Values |
|---|
| rdf:type
| |
| Description
| - The progress in computer development makes it possible to solve many problems of physical geodesy just by means of numerical arrangements unimaginable before. Nevertheless, considering a precise evaluation of spheroidal (spherical and ellipsoidal) harmonic functions in our typical task, we still observe a huge performance gap between numerical demands and capabilities of the present CPUs. Known methods for evaluation of Legendre functions (of the first as well as 2nd kind) are basically recurrent and thus sequential, where one step involves just a few arithmetic operations. Therefore, it seems to be challenging, but still feasible, to arrange an evaluation of Legendre functions in a way that admits wide SIMD parallelism. Regarding this aim, we provide Open Computing Language (OpenCL) implementation of the streaming-parallel algorithm for evaluation of ellipsoidal harmonic functions and their derivatives. The developed algorithm is free of assumptions about the function arguments, maximal degree/order or number of points to be evaluated and can be utilized on any vector data types of float, double or integer numbers. Besides, it addresses also floating-point issues in Legendre functions numerical treatment. We demonstrate the implementation on a GPGPU hardware, where anyone can easily obtain a computational power of several TFlops today. Several benchmarks are provided, including EGM2008, leading to conclusion about significant speedup of a single GPGPU over multi-core CPU device by a factor of 100, without any accuracy penalty. Moreover, thanks to OpenCL standard, we can benefit from excellent portability and scalability over heterogeneous parallel platforms, involving also hybrid MPI-OpenCL implementations. The topic presented is, of course, matter of importance in many application fields, not only in physical geodesy.
- The progress in computer development makes it possible to solve many problems of physical geodesy just by means of numerical arrangements unimaginable before. Nevertheless, considering a precise evaluation of spheroidal (spherical and ellipsoidal) harmonic functions in our typical task, we still observe a huge performance gap between numerical demands and capabilities of the present CPUs. Known methods for evaluation of Legendre functions (of the first as well as 2nd kind) are basically recurrent and thus sequential, where one step involves just a few arithmetic operations. Therefore, it seems to be challenging, but still feasible, to arrange an evaluation of Legendre functions in a way that admits wide SIMD parallelism. Regarding this aim, we provide Open Computing Language (OpenCL) implementation of the streaming-parallel algorithm for evaluation of ellipsoidal harmonic functions and their derivatives. The developed algorithm is free of assumptions about the function arguments, maximal degree/order or number of points to be evaluated and can be utilized on any vector data types of float, double or integer numbers. Besides, it addresses also floating-point issues in Legendre functions numerical treatment. We demonstrate the implementation on a GPGPU hardware, where anyone can easily obtain a computational power of several TFlops today. Several benchmarks are provided, including EGM2008, leading to conclusion about significant speedup of a single GPGPU over multi-core CPU device by a factor of 100, without any accuracy penalty. Moreover, thanks to OpenCL standard, we can benefit from excellent portability and scalability over heterogeneous parallel platforms, involving also hybrid MPI-OpenCL implementations. The topic presented is, of course, matter of importance in many application fields, not only in physical geodesy. (en)
|
| Title
| - An OpenCL implementation of ellipsoidal harmonics
- An OpenCL implementation of ellipsoidal harmonics (en)
|
| skos:prefLabel
| - An OpenCL implementation of ellipsoidal harmonics
- An OpenCL implementation of ellipsoidal harmonics (en)
|
| skos:notation
| - RIV/00025615:_____/13:#0001899!RIV14-MSM-00025615
|
| http://linked.open...avai/predkladatel
| |
| http://linked.open...avai/riv/aktivita
| |
| http://linked.open...avai/riv/aktivity
| |
| http://linked.open...vai/riv/dodaniDat
| |
| http://linked.open...aciTvurceVysledku
| |
| http://linked.open.../riv/druhVysledku
| |
| http://linked.open...iv/duvernostUdaju
| |
| http://linked.open...titaPredkladatele
| |
| http://linked.open...dnocenehoVysledku
| |
| http://linked.open...ai/riv/idVysledku
| - RIV/00025615:_____/13:#0001899
|
| http://linked.open...riv/jazykVysledku
| |
| http://linked.open.../riv/klicovaSlova
| - evaluation of Legendre functions; recurrent relations; SIMD parallelism; floating-point issues; GPGPU hardware (en)
|
| http://linked.open.../riv/klicoveSlovo
| |
| http://linked.open...i/riv/kodPristupu
| |
| http://linked.open...ontrolniKodProRIV
| |
| http://linked.open...i/riv/mistoVydani
| |
| http://linked.open...telVyzkumneZpravy
| - International Assocoation of Geodesy
|
| http://linked.open...in/vavai/riv/obor
| |
| http://linked.open...ichTvurcuVysledku
| |
| http://linked.open...cetTvurcuVysledku
| |
| http://linked.open...vavai/riv/projekt
| |
| http://linked.open...UplatneniVysledku
| |
| http://linked.open...iv/tvurceVysledku
| - Holota, Petr
- Nesvadba, Otakar
|
![[RDF Data]](/fct/images/sw-rdf-blue.png)
![[cxml]](/fct/images/cxml_doc.png)
![[csv]](/fct/images/csv_doc.png)
![[text]](/fct/images/ntriples_doc.png)
![[turtle]](/fct/images/n3turtle_doc.png)
![[ld+json]](/fct/images/jsonld_doc.png)
![[rdf+json]](/fct/images/json_doc.png)
![[rdf+xml]](/fct/images/xml_doc.png)
![[atom+xml]](/fct/images/atom_doc.png)
![[html]](/fct/images/html_doc.png)