COMPUTATIONAL TOOLKIT FOR NUMERICAL RELATIVITY
CAUTION: This page is still under construction !!
- The Computational Toolkit
- Toolkit Components:
- Finite-Difference Code "Skeletonizer"
- Support for Parallel Adaptive Finite-Difference Methods
- DAGH: A Data-Management Infrastructure for Parallel Adaptive Techniques
- User's Guide
- Draft Report
- Distribution (Version 0.1)
- DAGH Wish List
- DAGH Specification Requirements and Applications
Parallel Multigrid with the DAGH package: Specifications and Applications:
- RNPL: Rapid Numerical Prototyping Language (R. Marsa and M. Choptuik)
- The RNPL Reference Manual:
- The RNPL User's Guide:
- Man pages for BBHUTIL routines:
- The RNPL
html documentation (Cornell)
- Toolkit Meeting/Presentations:
- Computational Toolkit Meeting, Austin, TX (01/27/95)
- Presentations, (GC Meeting, Illinois, Nov'94)
- Jim Browne
- Matt Choptuik
- Manish Parashar
- GridSim: A Simulation Tool for Distributed Adaptive Grid Hierarchies (DAGHs)
- Fortran 90 Data Structure Specification (UT Austin + NPAC, Syracuse)
- HDF File Transfer Utilities
- Multigrid Bibliography (S. Klasky, M. Choptuik)
- Locally Developed Explorer (TM) Modules (M. Choptuik)
The objective of the computational toolkit is to provide language &
data-structure support, and a comprehensive development and execution
environment for numerical relativity applications on massively parallel
The project is part of the Binary Black Holes Grand Challenge
- Design Overview
(Robert Marsa & Matt Choptuik)
Phase I of the toolkit aims at developing a FD "skeletonizer" that
generates skeletons codes from an abstracts problem specification.
The user then fills in the code templates to complete an implementation
of the problem. Slides from Matt Choptuik's presentation summarizing
the Phase I implementation of the toolkit (Grand Challenge Meeting at
Urbana, Illinois, Nov 7-8) included below.
Phase I is largely completed thanks to Robert Marsa's implementation
of RNPL (view the reference manual),
a language for facilitating time-dependent finite-difference computations.
- Matt's Illinois Presentation
(Manish Parashar & Jim Browne)
Support for Parallel Adaptive Finite Difference Methods
We are currently developing programming abstractions and data-structure
support for parallel/distributed adaptive finite-difference methods.
A overview of issues and requirements as well as some design ideas were
presented at the Grand Challenge Meeting at Urbana, Illinois, (Nov 7-8).
A link to my tranperencies is included below. An early version (draft)
of a report describing our approach is also linked.
- A survey of software systems supporting parallel/distributed grids
- Illinois Presentation
Infrastructure for Parallel Adaptive Mesh-Refinement Techniques" [Draft]
- "DAGH: A Users Guide"
Partitioning Dynamic Adaptive Grid Hierarchies"
Distributed Dynamic Data-Structures for Parallel Adaptive Mesh-Refinement"
(Manish Parashar, Scott Klasky & Matt Choptuik)
GridSim is a trace-based simulator that provides the ability to visualize
the dynamic of the grid hierarchy associated with an application.
In addition it provides information about the computational load associated
with each grid, in inter-grid prolongation/restriction costs and the
regridding overheads. The final objective of this project is to provide a
platform to allow a developer to experiment with different distributions
of the grid hierarchy (DAGH) and analyse associated computational and
communication costs, and dynamic load balancing requirements.
- Official Fortran 90 data structure (May 1994)
- Example Fortan 90 program (Tom Haupt)
- Issues to be discussed (Tom Haupt)
- Summary of changes (Tom Haupt)
- Texas adaptation (Matt Choptuik)
- F77 3-d grid function (GF3) <--> HDF utilities
- Sample writer (creates the following HDF files:)
- Sample reader
- Utility routines for sample writer/reader
- f77 header file for development of HDF utilities
- Miscellaneous HDF Documentation
Supported by NSF ACS/PHY 9318152 (ARPA supplemented)
Manish Parashar, firstname.lastname@example.org
Center for Relativity & Department of Computer Sciences,
University of Texas at Austin