Bronis R. de Supinski Center for Applied Scientific Computing Lawrence Livermore National Laboratory June 2, 2005 - PowerPoint PPT Presentation

About This Presentation
Title:

Bronis R. de Supinski Center for Applied Scientific Computing Lawrence Livermore National Laboratory June 2, 2005

Description:

What makes new features worthwhile? Support for underlying OpenMP philosophy ... Work performed under the auspices of the U. S. Department of Energy by ... – PowerPoint PPT presentation

Number of Views:22
Avg rating:3.0/5.0
Slides: 7
Provided by: bron68
Category:

less

Transcript and Presenter's Notes

Title: Bronis R. de Supinski Center for Applied Scientific Computing Lawrence Livermore National Laboratory June 2, 2005


1
Bronis R. de SupinskiCenter for Applied
Scientific ComputingLawrence Livermore National
Laboratory June 2, 2005
The Most Needed Feature(s)for OpenMP 3.0
2
What makes new features worthwhile?
  • Support for underlying OpenMP philosophy
  • High-level constructs
  • Easy-to-understand semantics
  • Low cost extension/modification of serial code
  • Programmer needs
  • Variety of parallelism models
  • Control when appropriate
  • Clarity of specification (incontrovertible
    definitions)
  • Portability
  • Not just quality of implementation

3
Features for Clarity of Specification
  • Little details that 2.5 deferred to 3.0
  • Over a dozen small outstanding issues
  • Many concern clarity of specification(e.g.,
    directive grammar)
  • Orthogonality of constructs and base language
  • Reduction operators min and max
  • Array reductions
  • Allow unsigned integers as LCVs
  • (More) formalized memory model
  • Avoids natural language interpretations
  • Stated strictly in terms of operation orderings
  • Wont happen in 3.0

4
Several Worthy Candidates
  • Data distribution
  • Associate data to threads
  • Is this implementation- or architecture-specific?
  • Task queues
  • Supports a very common form of parallelism
  • Long-standing, well-understood proposals
  • Informational interface for tool support
  • Variable name mangling
  • Outline routines (or indicate that they arent
    used)
  • Run-time library names
  • Others

5
Personal Favorites
  • Contexts or subteams
  • Allow (subsets of) team to be reordered and named
  • Provides greater user control
  • Synchronization
  • Sections with varying parallelism
  • Supports portable libraries
  • Work distribution
  • User knows which thread should execute which work
  • Let them specify it!
  • Schedule rules help but not always natural
  • Often what users mean by data distribution

6
UCRL-PRES-212612
Work performed under the auspices of the U. S.
Department of Energy by University of California
Lawrence Livermore National Laboratory under
Contract W-7405-Eng-48
Write a Comment
User Comments (0)
About PowerShow.com