4th CERN Workshop on Numerical Computing

Europe/Zurich
Building 593, Room 11 (CERN)

Building 593, Room 11

CERN

Description

The Workshop

CERN openlab is pleased to announce a fourth instance of the Numerical Computing workshop, scheduled for May 5th and 6th 2014 at CERN. The course will be taught by external experts and several CERN staff. The workshop is targeted at attendees with good programming experience who would like to understand and control the floating-point calculations in their scientific programs.

In the 4th edition of the annual CERN Numerical Computing workshop we will be focusing more on vectorization and the efficiency of floating point. Again, we are proud to welcome world-class experts as speakers:
  • Florent de Dinechin will cover the metalibm package and FPGA floating point,
  • Agner Fog will speak about vector classes and instruction throughput measurements (w/ hands-on labs),
  • Danilo Piparo will introduce HEP floating point computing and the VDT library,
  • Lorenzo Moneta will floating point cover issues related to data analysis,
  • Jeff Arnold will provide a solid base and refresher for floating-point computing on modern x86 and with the GCC/ICC compilers (w/ labs).
Attendance is free, as for the regular openlab workshops, but but requires registration in EDH. If you don’t have EDH access, please let us know by e-mail that you’re planning to participate (openlab.workshops@cern.ch). A laptop is needed for accessing the servers that will be used for the exercises.
 
 
    • 09:15 10:00
      Introduction to floating point in Experimental HEP data processing and VDT 45m
      Speaker: Danilo Piparo (CERN)
      Slides
    • 10:00 10:45
      Floating Point issues in Data Analysis 45m
      Speaker: Lorenzo Moneta (CERN)
      Slides
    • 10:45 11:05
      Break 20m
    • 11:05 12:15
      Understanding FP arithmetic + hands-on 1h 10m
      Speaker: Jeffrey Arnold (Intel Corporation (US))
      Slides
    • 12:30 14:00
      Lunch (on your own) 1h 30m
    • 14:00 14:45
      FP arithmetic techniques 45m
      Speaker: Jeffrey Arnold
      Slides
    • 14:45 15:05
      Break 20m
    • 15:05 16:30
      Labs/Demos 1h 25m
      Speaker: Jeffrey Arnold (Intel Corporation (US))
      Slides
    • 09:00 10:00
      Methods for vectorization of mathematical software and performance measurements 1h
      Speaker: Prof. Agner Fog (Technical University of Denmark)
      Slides
    • 10:00 10:15
      Break 15m
    • 10:15 11:15
      Discussion + Lab 1h
      Speaker: Prof. Agner Fog
      notes
    • 11:15 11:30
      Break 15m
    • 11:30 12:20
      Metalibm updates 50m
      Speaker: Florent de Dinechin (ENS Lyon)
      Slides
    • 12:20 14:00
      Lunch (on your own) 1h 40m
    • 14:00 15:00
      Compiler support for floating point 1h
      Speaker: Jeffrey Arnold (Intel Corporation (US))
      Slides
    • 15:00 15:20
      Break 20m
    • 15:20 15:50
      Floating Point on FPGAs 30m
      Speaker: Florent de Dinechin (ENS Lyon)
      Slides
    • 15:50 17:00
      Compiler Lab 1h 10m
      Speaker: Jeffrey Arnold (Intel Corporation (US))