We are pleased to announce the fifth release (code name "Lovelace") of the Einstein Toolkit, an open, community developed software infrastructure for relativistic astrophysics. This release includes beginning support for OpenCL (disabled by default). In addition, bug fixes accumulated since the previous release in October 2011 have been included.
The Einstein Toolkit is a collection of software components and tools for simulating and analyzing general relativistic astrophysical systems that builds on numerous software efforts in the numerical relativity community including CactusEinstein, the Carpet AMR infrastructure and the relativistic hydrodynamics code GRHydro (an updated and extended version of the public release of the Whisky code). The Cactus Framework is used as the underlying computational infrastructure providing large-scale parallelization, general computational components, and a model for collaborative, portable code development. The toolkit includes modules to build complete codes for simulating black hole spacetimes as well as systems governed by relativistic hydrodynamics.
The Einstein Toolkit uses a distributed software model and its different modules are developed, distributed, and supported either by the core team of Einstein Toolkit Maintainers, or by individual groups. Where modules are provided by external groups, the Einstein Toolkit Maintainers provide quality control for modules for inclusion in the toolkit and help coordinate support. The Einstein Toolkit Maintainers currently involve postdocs and faculty from five different institutions, and host weekly meetings that are open for anyone to join in.
Guiding principles for the design and implementation of the toolkit include: open, community-driven software development; well thought out and stable interfaces; separation of physics software from computational science infrastructure; provision of complete working production code; training and education for a new generation of researchers.
For more information about using or contributing to the Einstein Toolkit, or to join the Einstein Toolkit Consortium, please visit our web pages at <http://einsteintoolkit.org>.
The Einstein Toolkit is primarily supported by NSF 0903973/0903782/0904015 (CIGR), and also by NSF 0701566/0855892 (XiRel), 0721915 (Alpaca), 0905046/0941653 (PetaCactus), and 0710874 (LONI Grid).
The Einstein Toolkit contain over 170 regression test cases. On a large portion of the tested machines, all of these testsuites pass, using both MPI and OpenMP parallelisation.
This release of the Einstein Toolkit adds support for GPUs and other accelerators. This support comprises three levels of abstraction, ranging from merely building and running both CUDA and OpenCL code, to automated code generation targeting GPUs instead of CPUs. As with any other programming paradigm (such as MPI or OpenMP), the performance benefits depend on the particular algorithms used and optimizations that are applied. In addition, the Simulation Factory greatly aids portability to a wide range of computing systems.
At the lowest level, Cactus now supports compiling, building, and running with either CUDA or OpenCL. CUDA is supported as new language in addition to C, C++, and Fortran; OpenCL is supported as an external library, and builds and executes compute kernels via run-time calls. Details are described in the user's guide (for CUDA) and in thorn ExternalLibraries/OpenCL (for OpenCL).
Many accelerator platforms today separate between host memory and device memory, and require explicit copy or map operations to transfer data. An intermediate level of abstraction aids transferring grid variables between host and device, using schedule declarations to keep track of which data are needed where, and minimizing expensive data transfers. For OpenCL, there is a compact API to build and execute compute kernels at run time. Details are described in thorns CactusUtils/Accelerator and CactusUtils/OpenCLRunTime (with example parameter file).
Finally, the code generation system Kranc has been extended to be able to produce either C++ or OpenCL code, based on the infrastructure described above. This allows writing GPU code in a very high-level manner. However, it needs to be stated that the efficiency of the generated code depends on many variables, including e.g. the finite differencing stencil radius and the number of operations in the generated compute kernels. Non-trivial kernels typically require system-dependent tuning to execute efficiently, as GPUs and other accelerators generally show a rather unforgiving performance behavior. The thorns McLachlan/ML_WaveToy and McLachlan/ML_WaveToy_CL are examples, generated from the same Kranc script, showing the generated C++ and OpenCL code.
To upgrade from the previous release, use GetComponents with the new thornlist to check out the new version. GetComponents can also be used to update an existing checkout, but since some components might have changed location or have been removed from the toolkit you should try a new checkout whenever possible instead.
See the Download page on the Einstein Toolkit website for download instructions.
All repositories participating in this release carry a branch ET_2012_05 marking this release. These release branches will be updated if severe errors are found.
This release has been tested on about twenty systems and architectures, including the following:
The Simulation Factory contains ready-to-use configuration details for more than 40 additional systems, including most HPC systems at LONI, NERSC, RZG, and XSEDE (TeraGrid).
The Einstein Toolkit web site contains online documentation for its components, and pointers for using it to build your own code. There is also a tutorial that explains how to download, build, and run the code for a simple binary black hole evolution. We invite you to join our mailing list <users@einsteintoolkit.org>.
On behalf of the Einstein Toolkit Consortium: the "Lovelace" Release Team
Eloisa Bentivegna
Tanja Bode
Peter Diener
Roland Haas
Ian Hinder
Frank Löffler
Bruno Mundim
Christian D. Ott
Erik Schnetter
May 28, 2012