Accelerating the 'fields' package: Using parallel linear algebra software [presentation]
Lyngaas, I., Paige, J., Vadlamani, S., Nychka, D.. (2014). Accelerating the 'fields' package: Using parallel linear algebra software [presentation].
| Title | Accelerating the 'fields' package: Using parallel linear algebra software [presentation] |
|---|---|
| Genre | Conference Material |
| Author(s) | Isaac Lyngaas, John Paige, Srinath Vadlamani, Douglas Nychka |
| Abstract | Statistical packages such as the 'Fields' package in R can benefit from using multicore systems and accelerated hardware such as GPUs and Xeon Phis. Performance of linear algebra operations such as those used in spatial statistics computations in 'Fields' can be accelerated by new accelerator hardware. A survey is completed for several multicore and accelerator based linear algebra packages, including the commercial package CULA, and PLASMA and MAGMA both from the University of Tennessee, to determine the possible speedup. For double precision matrices of size N=5000 or greater, we find that computing a Cholesky decomposition using CULA we can get at least an 18 times speedup over conventional R. This method uses a single Nvidia Tesla M270Q GPU within Caldera. |
| Publication Title | |
| Publication Date | Jul 31, 2014 |
| Publisher's Version of Record | |
| OpenSky Citable URL | https://n2t.org/ark:/85065/d72z175x |
| OpenSky Listing | View on OpenSky |
| CISL Affiliations | CISLAODEPT, TDD, ASAP |