Note: The BEARCLAW website is still under construction. An updated distribution is available from this server (Nov. 15, 2001). A few bugs have been resolved. The wave propagation schemes and classic schemes are stable. Other solver modules are not yet robust enough for public distribution. Features are annotated as:
stable - feature works on all problems, the code may still crash but this is highly likely to be due to the coding of the physical problem being solved
functional - feature works on most problems, but has been observed to cause crashes not related to the physical problem being solved
experimental - feature works on a few, simplified problems; frequent crashes not related to the physical problem being solved.
not implemented - planned feature
Stable features are included in the public distribution. Functional features are available upon request.
Week of Jan. 7, 2002 - Flexible shock tube problem, plot of Mach field
GIF animation (1.4 MB), More details
Previous showcase computations.
BEARCLAW is a general purpose software package for solving time dependent partial differential equations.
Main features:
Automatic adaptive mesh refinement
Prior to each time step a predictor step on a coarser grid is used to estimate errors. Finer grids are generated in accordance with a specified criterion in regions of higher error (stable)
A posteriori error estimator that does not require coarse grid step (experimental)
Parallel execution
Distributed memory paradigm, MPI implementation (functional)
Most computations are programmed as Fortran 90 vector operation thus exposing fine-grained parallelism for vector machines (stable)
Multi-physics capability
Multiple PDE's may be solved simultaneously by a priori association of a subdomain with the PDE-specific routines (functional)
Dynamic association of subdomains with PDE-specific routines allows computation of media with changing constitutive properties (functional)
Dynamic data structures
Memory for the PDE fields is allocated dynamically, as needed (stable)
A variety of output formats
AMRCLAW format (stable)
TECPLOT format (stable)
HDF format (stable)
Unified code base for 1-4 dimensional PDE's (stable)
Encapsulation of all PDE specific data in a single structure which may be modified to suit particular needs (stable)
Boundary embedding for computation in domains with complicated geometry (not implemented)
Fortan 90 compiler. Code has been developed and tested with the NAG f95 compiler on Linux platforms and the Digital f90 compiler on Alpha processor platforms running Unix
HDF (Hierarchical Data Format) libraries. Development and testing has been carried out with the HDF libraries v4.3 and v4.4
The BEARCLAW documentation (gzip compressed Postscript files):
User's Manual - describes how to set up a BEARCLAW application without undue detail of the inner workings (40 % complete)
Programmer's Manual - fully describes BEARCLAW, shows how additional solver modules may be implemented (5% complete)
Unpack the following tar archive in your local $CLAW directory using:
tar zxvf bearclaw.tar.gz
The archive contains stable BEARCLAW code. If you're updating a previously installed version remember to save your $BEARCLAW/Makefile.inc file that contains system specific settings (compiler switches, library paths).
You have two options for using BEARCLAW:
Create a local BEARCLAW directory under your $CLAW directory and unpack the distribution as stated above (Here's a Makefile.inc file suited to the AMath system).
Use the precompiled library files directly from the main $BEARCLAW directory. In this case you would set your $BEARCLAW environment variable as:
setenv BEARCLAW /home/claw/bearclaw
In your home directory you would only have the application files, i.e. nodeinfo.f90 and problem.f90, and any data files you need.
The second technique is recommended unless you're interested in modifying the BEARCLAW library files. It gives you immediate access to the latest version of the software.
Fairly simple applications of BEARCLAW. Useful as a tutorial introduction and a quick overview of basic functionality.
More complex applications of BEARCLAW. Shows multiphysics capabilities, parallel execution, embedded boundaries.
