Here is our current list of DxT publications. The most recent papers are listed first. See the copyright notice.
To search for multiple keywords separate each keyword
by a comma: keyword , keyword
Keywords include:
architectures, analysis, aspects, categories, commuting diagrams, computational design, derivatives, Dense Linear Algebra (DLA),
Design by Transformation (DxT), design rules, Design Rule Checking (DRC), evolution, feature interactions, feature models, geodesic,
higher-order transformations, kubes, MDA, MDD, MDE, mixin layers, optimization, optional feature problem, origami, refactoring,
streaming, safe composition, semantics, testing, UML, verification, wizards, P3 (P1/P2/P3/DiSTiL data structure generator).
B. Marker, D. Batory, and R. van de Geijn. A Case Study in Mechanically Deriving Dense Linear Algebra Code . To appear in International Journal of High Performance Computing Applications.
Design by Transformation (DxT) is a top-down approach to mechanically derive high-performance algorithms for dense linear algebra. We use DxT to derive the implementation of a representative matrix operation, two- sided Trmm. We start with a knowledge base of transformations that were encoded for a simpler set of operations, the level-3 BLAS, and add only a few transformations to accommodate the more complex two-sided Trmm. These additions explode the search space of our prototype system, DxTer, requiring the novel techniques defined in this paper to eliminate large segments of the search space that contain suboptimal algorithms. Performance results for the mechanically optimized implementations on 8,192 cores of a BlueGene/P architecture are given.
DxT, DLA, MDE, MDA, MDD
B. Marker, D. Batory, R. van de Geijn. DSLs, DLA, DxT, and MDE in CSE . International Workshop on Software Engineering for Computational Science and Engineering (SECSE), May 2013.
We narrate insights from a collaboration between researchers in Software Engineering (SE) and in the domain of Dense Linear Algebra (DLA) libraries. We highlight our impressions of how software development for computational science has traditionally been different from the development of software in other domains. We observe that scientific software (at least DLA libraries) is often developed by domain experts rather than legions of programmers. For this reason, researchers in SE need to impact the productivity of experts rather than the productivity of the masses. We document this and other lessons learned.
DSLs, DxT, DLA, MDE, MDA, MDD
B. Marker, D. Batory, R. van de Geijn. Code Generation and Optimization of Distributed-Memory Dense Linear Algebra Kernels . International Workshop on Automatic Performance Tuning (IWAPT), June 2013.
Design by Transformation (DxT) is an approach to software development that encodes domain-specific programs as graphs and expert design knowledge as graph transformations. The goal of DxT is to mechanize the generation of highly-optimized code. This paper demonstrates how DxT can be used to transform sequential specifications of an important set of Dense Linear Algebra (DLA) kernels, the level-3 Basic Linear Algebra Subprograms (BLAS3), into high-performing library routines targeting distributed-memory (cluster) architectures. Getting good BLAS3 performance for such platforms requires deep domain knowledge, so their implementations are manually coded by experts. Unfortunately, there are few such experts and developing the full variety of BLAS3 implementations takes a lot of repetitive e ort. A prototype tool, DxTer, automates this tedious task. We explain how we build on previous work to represent loops and multiple loop-based algorithms in DxTer. Performance results on a BlueGene/P parallel supercomputer show that the generated code meets or beats implementations that are hand-coded by a human expert and outperforms the widely used ScaLAPACK library.
DxT, DLA, MDE, MDA, MDD, optimization, streaming
T. Riché, R. Goncalves, B. Marker, and D. Batory. Pushouts in Software Architecture Design . Generative Programming and Component Engineering (GPCE), September 2012.
A classical approach to program derivation is to progressively extend a simple specification and then incrementally refine it to an implementation. We claim this approach is hard or impractical when reverse engineering legacy software architectures. We present a case study that shows optimizations and pushouts -- in addition to refinements and extensions -- are essential for practical stepwise development of complex software architectures.
architectures, DxT, MDE, MDA, MDD, riche
B. Marker, J. Poulson, D. Batory, and R. van de Geign. Designing Linear Algebra Algorithms by Transformation: Mechanizing the Expert Developer . International Workshop on Automatic Performance Tuning (IWAPT), July 2012.
To implement dense linear algebra algorithms for distributed-memory computers, an expert applies knowledge of the domain, the target architecture, and how to parallelize common operations. This is often a rote process that becomes tedious for a large collection of algorithms. We have developed a way to encode this expert knowledge such that it can be applied by a system to generate mechanically the same (and sometimes better) highly-optimized code that an expert creates by hand. This paper illustrates how we have encoded a subset of this knowledge and how our system applies it and searches a space of generated implementations automatically.
architectures, DxT, MDE, MDA, MDD.
J. Feigenspan, D. Batory, and T. Riché. Is the Derivation of a Model Easier to Understand than the Model Itself? . International Conference on Program Comprehension (ICPC), June 2012.
Software architectures can be presented by graphs with components as nodes and connectors as edges. These graphs, or models, typically encode expert domain knowledge, which makes them difficult to understand. Hence, instead of presenting a complete complex model, we can derive it from a simple, easy-to-understand model by a set of easy-to-understand transformations. In two controlled experiments, we evaluate whether a derivation of a model is easier to understand than the model itself.
architectures, DxT, MDE, MDA, MDD, riche
B. Marker, A. Terrel, J. Poulson, R. van de Geijn, and D. Batory. Mechanizing the Expert Dense Linear Algebra Developer (Poster Paper) . Principles and Practice of Parallel Programming (PPoPP), February 2012.
The efforts of an expert to parallelize and optimize a dense linear algebra algorithm for distributed-memory targets are largely mechanical and repetitive. We demonstrate that these efforts can be encoded and automatically applied to obviate the manual implementation of many algorithms in high-performance code.
DxT, MDE, DLA, dense linear algebra
B. Marker, D. Batory, J. Poulson, and R. van de Geijn. How a Domain-Specific Language Enables the Automation of Optimized Code for Dense Linear Algebra (Extended Abstract) . International Workshop on Domain-Specific Languages and High-Level Frameworks for High Performance Computing (WOLFHPC), May 2011.
DxT, MDE, MDA, MDE, streaming architectures, DLA
T.L. Riché, H.M. Vin, and D. Batory. Transformation-Based Parallelization of Request-Processing Applications . Model Driven Engineering Languages and Systems (MODELS), October 2010.
Multicore, multithreaded processors are rapidly becoming the platform of choice for high-throughput request-processing applications (RPAs). We refer to this class of modern parallel platforms as multi-* systems. In this paper, we describe the design and implementation of Lagniappe, a translator that simplifies RPA development by transforming portable models of RPAs to highthroughput multi-* executables. We demonstrate Lagniappe’s effectiveness with a pair of stateful RPA case studies.
MDD riche MDE computational design award streaming architectures DxT
D. Batory (work with T. Riché). Refinement and Optimization of Streaming Architectures. Keynote at Software Engineering and Data Engineering (SEDE), Las Vegas, June 2009 and the Software Engineering and Databases (JISDB), San Sebastian, Spain, September 2009.
We present a Model Driven Engineering approach to explain, verify,
build, and test dataflow or streaming software architectures that
are to be parallelized for performance or availability. Componentconnector
models are incrementally elaborated by transformations
that refine or optimize architectural designs. We re-engineered two
significant case studies to illustrate the generality of our work: (1)
recoverable crash fault tolerant servers and (2) join parallelizations
in database machines.
PDF of
presentation
DxT, keynote, pipe-and-filter, architectural refinement, optimization, riche
Copyright Notice: This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder.