*************************************************************************** First HPCnet workshop on: European Toolsets for High Performance Computing *************************************************************************** High Performance Computing has become a cost-effective solution for the ever growing need for computing power in research and industry. The introduction of this new technique into the present set of applications and the efficient exploitation thereof, needs to be supported by a set of tools that will aid the migration to and acceptance of HPC technology. This need for tool support has been recognized by many groups. Within the framework of a number of European (Esprit) projects, many tools have been developed. They include performance evaluators (simulators and monitors), porting utilities (interactive performance prediction and libraries), complete parallel programming environments, and tools for efficient resource utilization. The goal of this workshop is to present an overview of European toolsets for HPC. The emphasis will be on employment of the tools and related practical aspects. This workshop is especially directed at (potential) Industrial users of HPC technology. *************************************************************************** Date: Thursday, April 13th, 1995 Organization: HPCnet, University of Amsterdam Location: Amsterdam Programme: ---------- 9:00 Coffee 9:30 Welcome and introduction Dr P. M. A. Sloot, University of Amsterdam 9:45 The PREPARE HPF Programming Environment A. H. Veen, ACE Associated Computer Experts BV 10:45 The PPPE toolset Prof. Dr A. J. G. Hey, University of Southampton 11:45 Demonstration of the PPPE toolset 13:00 Lunch 14:00 The CAMAS toolset Dr J. Reeve, University of Southampton 15:00 The CODINE and TRAPPER toolsets Prof. Dr W. Gentzsch, Genias 16:00 Demonstration of the CAMAS, CODINE, and TRAPPER toolsets 16:30 Refreshments, while the demonstration continues. *************************************************************************** Abstracts and tools descriptions: The PREPARE HPF Programming Environment A. H. Veen ACE Associated Computer Experts BV --------------------------------------- The Esprit project PREPARE develops a programming environment in which High Performance Fortran (HPF) programs can be developed or restructured in a machine-independent fashion and executed on a massively parallel MIMD machine. HPF is an extension of Fortran 90 containing a set of compiler directives that specify how arrays are to be distributed over a set of processors. The PREPARE environment is based on: - a Parallelization Module that transforms the original data-parallel form of the source program into Single Program Multiple Data (SPMD) form, - an Interactive Engine that provides performance measures, and - an efficient Compilation System that generates highly optimized code The unique strength of the PREPARE project is the tight integration of these three aspects. The compilation system and the parallelization module are linked through a powerful interface that enables the use of each other's analysis information and mutually influence each other's optimization decisions. Portable Parallel Programming Environments The ESPRIT PPPE Project A. J. G. Hey, University of Southampton ------------------------------------------ There is a strong belief in the potential of large distributed memory parallel computers in the development of large scale scientific and engineering applications. However, several factors currently prevent the widespread uptake of these machines within an industrial and commercial context. Chief among these are the lack of integrated tool environments, lack of portability across parallel architectures, lack of migration aids, and lack of adherence to Open System standards. Under the auspices of the CEC ESPRIT 3 Initiative, European hardware manufacturers together with software houses and large industrial organizations have come together to form a consortium developing a fully integrated set of parallel programming tools. PPPE aims to produce a portable, parallel programming environment available across many high performance and parallel computers such that these systems become available for mainstream applications development. Very broadly there are two sides to the environment: offering run-time and the cross-development functionality. Each side has a portability layer upon which all the components are based: PCTE for the cross-development environment and PARMACS/MPI for the run-time environment. The initial tools being developed include a parallel debugger, a performance monitor and analyzer and a HPF mapper and debugger. The CAMAS toolset: ------------------ Unfortunately, an abstract of the CAMAS toolset presentation is not available at the time of this announcement. Software Tools for Batch and Parallel Computing Wolfgang Gentzsch, GENIAS Software and FH Regensburg ---------------------------------------------------- In the lecture two topics of recent interest will be discussed: One is computing in heterogeneous clusters, and the other one is tool-supported development of parallel software. For illustration two representative European software tools will be presented in more detail: CODINE and TRAPPER: The software CODINE (Computing in Distributed Networked Environments) is a job management environment that distributes compute-intensive jobs over local and wide area networks to powerful HPC systems in a loadbalanced way. The central CODINE Manager sitting somewhere in a dedicated Centre collects all system specific information including system load, from all systems in the network, and according to the requirements of the Medical and of the HPC partners, the manager sends the jobs to the most suitable and least loaded HPC system. Thus a well-balanced use of all available systems is achieved and the user has the guarantee that his job is submitted to the optimal system. In addition some other CODINE features allow for easy and unified job administration such as monitoring and accounting of jobs. The software TRAPPER is a graphic programming environment for parallel systems. It has been extensively used e.g. at Daimler Benz to write complex parallel software for embedded parallel systems, e.g. in the Prometheus Project of a driverless automobile with cameras and parallel processors. The novel approach introduced with TRAPPER is to support the different stages of the software engineering process with emphasis on the specific problems of parallel systems. The programming environment contains components for software design,hardware configuration, mapping, monitoring, software visualization, and performance monitoring of parallel applications and systems. TRAPPER can be used for all software development in the Project thus guaranteeing the same homogeneous programming strategy and paradigm, among others allowing an easy exchange of software and the possibility of co-development among different group in the future. The HPCnet goals and objectives ------------------------------- HPCnet is an Esprit-funded project (project number 09004) to establish a pan-European network of excellence in high performance computing: a network of people with expertise and experience. The overall objective of HPCnet is to catalyse the exploitation of HPC technologies by European industry. This will be achieved by providing information through a Virtual Supermarket and by organizing focussed workshops. *************************************************************************** Location and route: Room Z011 CWI Kruislaan 413 1098 SJ Amsterdam The Netherlands Route available on WWW: http://www.cwi.nl/cwi/reaching_cwi.html The demonstration of the toolsets will be held in the same building. The exact location will be announced during the presentations on April 13th. *************************************************************************** Participation: Costs: there is no fee for attendance, yet there is no budget to provide lunch for the participants. The cafeteria at the location is adequately equipped for the expected guests. If you are planning to attend, it is important that you inform the us before Monday, March 20th, 1995. This information is also available on the HPCnet WWW server: http://hpcnet.soton.ac.uk/ To attend, please fill out this form and send it by email, fax or conventional mail to the workshop contact: Arjan de Mes Parallel Scientific Computing and Simulations Group Department of Mathematics and Computer Science University of Amsterdam Kruislaan 403 1098 SJ Amsterdam The Netherlands email: mes@fwi.uva.nl fax: +31 20 525 7490 Participation form: Last name: ............................................................ First name: ........................................................... Title: ................................................................ Organization: ......................................................... Phone: ................................................................ Fax: .................................................................. Email: ................................................................ Address: .............................................................. Zip-code: ............................................................. City: ................................................................. Country: .............................................................. ***************************************************************************