 |
| Titel |
A GPU Reaction Diffusion Soil-Microbial Model |
| VerfasserIn |
Ruth Falconer, Alasdair Houston, Sonja Schmidt, Wilfred Otten |
| Konferenz |
EGU General Assembly 2014
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 16 (2014) |
| Datensatznummer |
250098388
|
| Publikation (Nr.) |
 EGU/EGU2014-14059.pdf |
|
|
|
|
|
| Zusammenfassung |
| Parallelised algorithms are frequent in bioinformatics as a consequence of the close link to
informatics – however in the field of soil science and ecology they are less prevalent. A
current challenge in soil ecology is to link habitat structure to microbial dynamics. Soil
science is therefore entering the ‘big data’ paradigm as a consequence of integrating data
pertinent to the physical soil environment obtained via imaging and theoretical models
describing growth and development of microbial dynamics permitting accurate analyses of
spatio-temporal properties of different soil microenvironments. The microenvironment is
often captured by 3D imaging (CT tomography) which yields large datasets and when used in
computational studies the physical sizes of the samples that are amenable to computation are
less than 1 cm3. Today’s commodity graphics cards are programmable and possess
a data parallel architecture that in many cases is capable of out-performing the
CPU in terms of computational rates. The programmable aspect is achieved via a
low-level parallel programming language (CUDA, OpenCL and DirectX). We ported a
Soil-Microbial Model onto the GPU using the DirectX Compute API. We noted a
significant computational speed up as well as an increase in the physical size that can
be simulated. Some of the drawbacks of such an approach were concerned with
numerical precision and the steep learning curve associated with GPGPU technologies. |
|
|
|
|
|