JUNE 18–22, 2017
FRANKFURT AM MAIN, GERMANY

Session Details

 
Name: Panel: 65 Years of Compiler Development
 
Time: Monday, June 19, 2017
05:00 pm - 06:00 pm
 
Room:   Panorama 2
Messe Frankfurt
 
Moderator:   Georg Hager, RRZE
  Gerhard Wellein, RRZE & University of Erlangen-Nuremberg
 
Panelist:   Jan Eitzinger, RRZE
  Sebastian Hack, Universität des Saarlandes
  Paul Kelly, Imperial College London
  Harald Köstler, University of Erlangen-Nuremberg
  Louis-Noel Pouchet, Ohio State University
  Michael Wolfe, NVIDIA/PGI
 
Abstract:  
The compiler is the developer's dearest friend and most bitter enemy. It displays fabulous brilliance, blatant stupidity, and an annoying inability to read peoples' minds, all at the same time. "Let the compiler figure it out" is arguably the premier performance impediment in HPC, and from a pure developer's point of view the history of compiler development is abundant with unfulfilled expectations. On the other hand, there are those who think that the whole problem will go away if we change the way we talk to the system, either by adding more abstraction or by drilling down to the hardware.  Moreover, latest developments such as parallel languages, metacompilers, DSLs, and advanced loop transformations promise a dawn of mighty compiler magic.  This panel wants to bring together developers from both sides of the compiler to discuss whether dissatisfaction with its abilities is inevitable, what can be done about it, and what the future has in store. Controversial opinions are expected and welcome!