dens linear algebra librari need cope effici rang input problem size shape inherent mean parallel implement have exploit parallel wherev present while openmp allow relat fine grain parallel exploit share memori environ current lack featur make easi partit comput over multipl array indic overlap sequenti parallel comput the inher flexibl natur share memori paradigm openmp pose other difficulti when becom necessari optimis perform across success parallel librari call notion borrow from distribut memori paradigm explicit data distribut help address some problem focus data rather than work distribut appear misplac smp context

C. Addison

Y. Ren

M. van Waveren

OpenMP Issues Arising in the Development of Parallel BLAS and LAPACK Libraries

the openmp standard defin applicat program interfac api share memori comput sinc it introduct 1997 has grown becom one most common use api parallel program but success market doesn necessarili impli success comput scienc openmp good program environ what doe even mean call program environ good and final onc understand how good bad openmp what can make even better paper address question

Timothy G. Mattson

How Good is OpenMP

Eduard Ayguade

Barbara Chapman

Introduction

this paper describ design develop softwar packag support variabl precis arithmet semant extens fortran 95 languag the work precis arithmet support packag can dynam arbitrarili vari the facil exploit dataabstract capabl fortran 95 allow oper use element array operand well scalar the number system defin way close under all basic oper normal arithmet programtermin numer except can occur precis loss situat like underflow overflow handl defin special valu represent preserv much numer inform practic oper semant defin except valu propag appropri reflect loss inform the number system use essenti convent variabl precis floatingpoint represent when oper can perform exact within currentlyset work precis limit excess trail zero digit store nor take part futur oper this both econom storag improv effici

J.L. Schonfelder

Variable Precision Arithmetic: A Fortran 95 Module

java receiv increas attent most popular platform distribut comput howev programm still reluct embrac java tool write scientif engin applic due it still notic perform drawback compar other program languag fortran paper present hybrid java fortran implement parallel particleincel pic algorithm plasma simul our approach timeconsum compon applic design implement fortran subroutin while less calculationintens compon usual involv build user interfac written java the two type softwar modul have been glu togeth use java nativ interfac jni our mixedlanguag pic code test it perform compar pure java fortran version same algorithm sun e6500 smp system linux cluster pentium iii machin

Quanming Lu

Vladimir Getov

Mixed-Language High-Performance Computing for Plasma Simulations

symbol solver generat deal system partial differenti equat pdes function arbitrari number variabl present can also handl arbitrari domain geometri independ variabl given system pdes solver generat set explicit finitediffer method ani specifi order fourier stabil criterion each method for method stabl iter function generat symbol use pde it initi boundari condit this iter function dynam generat everi pde problem it evalu provid solut pde problem fortran 90 code iter function generat use mathcod system which result perform gain order thousand over mathematica languag has been use code solver generat exampl stabil criteria present agre known criteria exampl demonstr general solver speed enhanc generat fortran 90 code also present

K. Sheshadri

Peter Fritzson

A General Symbolic PDE Solver Generator: Explicit Schemes

this paper present flowsensit algorithm comput interprocedur definitionus chain dynam pointerlink data structur the goal relat statement construct link dynam pointerlink data structur definit statement might travers structur through link use specif each statement defin link pointerlink data structur algorithm find set statement travers link which defin this algorithm solv definitionus chain problem perform backward iter data flow analysi comput set upward expos use each statement the result algorithm can use identifi parallel program even cyclic pointerlink data structur

Yuan-Shin Hwang

Joel Saltz

Interprocedural Definition-Use Chains of Dynamic Pointer-Linked Data Structures

Ronald Perrott

Boleslaw Szymanski

Changes in the Editorial Board

About this journal

Scientific Programming provides a forum for research results in, and practical experience with, software engineering environments, tools, languages, and models of computation aimed specifically at supporting scientific and engineering computing.

Journal profile

Meet the editorial board

Chief Editor, Professor Tramontana, is based at the University of Catania and his research primarily concerns the areas of software engineering and distributed systems.

Editor spotlight

Submit to Special Issue

We currently have a number of Special Issues open for submission. Special Issues highlight emerging areas of research within a field, or provide a venue for a deeper investigation into an existing research area.

Scientific Programming