Fox Presentation Fall 1995 Scalable Scientific Software LIbraries and Problem Solving Environments Virtual Parallel Environments and Languages Purdue 25-27 September 1995 Moderator: Geoffrey Fox NPAC Syracuse University 111 College Place Syracuse NY 13244-4100 Some Overall Issues --I Need to define terms more precisely: Does Language refer to fine grain operations/parallelism as in HPF Does Environment refer to coarse grain parallelism as seen in AVS or Khoros? Is "Software Bus" same thing as coarse grain software integration? Does a scripting (interpreted) language such as PERL (an environment for document manipulation) or Visual Basic encompass both terms? Does Parallel include Distributed? What is "Virtual" ? What are requirements? At system level -- metacomputers, and/or MPP's, and/or PC's ... At Application level -- fields include Partial Differential Equations, Image Processing .. Distributed Simulation as in SIMNET .. DeskTop Publishing? Some Overall Issues -- II How hard should one try to build open systems allowing disparate paradigms? This is essential in say manufacturing where need to integrate 10,000 existing programs in Excel, Lisp .. Fortran together How important is ability to invoke general purpose languages (such as Fortran, CC++, HPF ..) as well as a script with domain specific constructs. cf: NSF Blackhole project with HPF and more focused approachs -- latter are ahead as HPF slow! General and most serious problem with all HPCC software -- how do we support system with hardware rapidly evolving and software (industry) support resources small Role of packages such as Mathematica cf: SINAPSE Some Example Systems with Ideas and Requirements MATLAB -- note not parallel but very succesful AVS -- different level of abstraction from MATLAB -- also succesful and not (correctly) parallel. PDElab/ELLPACK and other partial differential equation packages DSI / SIMNET and distributed event driven simulation for military applications VRML and growing distributing interactive gaming applications Global Arrays (Pacific Northwest Lab) -- very succesful toolkit for matrix formulated chemistry applications -- can they/should they extend to finer grain molecular dynamics ASOP -- Multidisciplinary Design Systems -- and lessons from prototypes Engineous (GE), FIDO(Langley), NPSS(Lewis) Note link to databases and need for wrappers for legacy systems WebWork -- Boston NPAC -- use Web for (initially) coarse grain software integration