BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20111117T013000Z
DTEND:20111117T030000Z
LOCATION:TCC LL4
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: Scripting languages such as Python and Javascript, and Domain-Specific Languages (DSLs) such as R hold out the promise of higher productivity application development.  However, the performance of scripting languages can be poor.  To enable the use of scripting languages for HPC, it is possible to embed a DSL in them, which allows reuse of existing infrastructure for tasks which are not performance critical. DSLs make the application developer’s task easier by providing programming models tuned for particular application areas.  By supporting more restrictive programming models, DSLs can target higher performance.  We will discuss NumPy/SciPy as well as Copperhead, both of which are embedded in Python; a data-parallel extension to Javascript which supports performance-critical applications embedded in web browsers; and a high-performance implementation of R.  NumPy/SciPy takes a library approach, while Copperhead compiles to NVIDIA’s CUDA and the high-performance R implementation uses Intel’s Array Building Blocks virtual machine.
SUMMARY:Scripting Languages and DSLs for Productive Development of HPC Applications
PRIORITY:3
END:VEVENT
END:VCALENDAR