The goal of this project is to develop a high-performance algorithmic and software framework for solving partial differential equations arising from problems in three important mission areas for the DOE Office of Science: magnetic fusion, accelerator design, and combustion. This framework will provide investigators in these areas with a new set of simulation capabilities based on locally structured grid methods, including adaptive meshes for problems with multiple length scales; embedded boundary and overset grid methods for complex geometries; efficient and accurate methods for particle and hybrid particle/mesh simulations; and high-performance implementations on distributed-memory multiprocessors. We are undertaking this project as an end-to-end process, with close interactions with stakeholders from these applications disciplines.
The project is organized around three component activities.
Applications Development. We use the software tools
developed in this project to implement specific target applications.
These are developed as joint projects with scientists
from the three communities listed above, and will be
applied to solve problems of independent scientific
interest in those areas. Individual Applications Development
projects would be scoped to take 1-2 years, and have
a direct impact on the Office of Science applications.
Software Development. We are developing the core software libraries
that serve as the components out of which the applications are
built. Our stakeholders have a common set of algorithmic and software
requirements, and we use
state-of-the-art software design methodologies to ensure reuse across
applications. We also leverage
de-facto standards in the community to the maximum extent possible,
including ones that are being developed under other parts of
SciDAC. As various software libraries are developed, they serve as the
basis for the next set of applications codes to be developed under the
Applications Development activity. Thus, a Software Development project
would be expected to to have an impact on Office of
Science applications in a 2-3 year time frame.
Algorithm Development. We are pursuing promising
ideas from the research community that have the potential to
become robust algorithms that can be incorporated
into the next generation of our software.
In this activity, we develop specifications of
algorithms that would constitute input to the Software Development
activity, with an impact on the Office of
Science applications in a 3-5 year time frame.
This proposal addresses in detail only LBNL's activities, which are in the first two areas.
This work provides the following major benefits to the DOE SciDAC program. First, it provides to the SciDAC applications users substantial new algorithmic capabilities not previously available in their simulation codes. Second, the cost of providing that capability is considerably less than if the applications stakeholders set out to do this on their own. The overlap between the requirements of these three applications areas is substantial, so that the redundant effort that would take place if the capabilities were developed independently is avoided. In addition, the PI's on this project have an accumulated expertise in developing algorithms and software in these areas that would be difficult for the applications areas to reproduce on a short time scale. Third, the close coupling to the applications in the development activities in this project serves to transfer expertise in these areas to the SciDAC applications scientists. Conversely, the involvement of some of the leading applied mathematicians in the DOE community in the solution of SciDAC problems will lead to new applied mathematics research that will ultimately benefit both communities.