| Version 1.9 Build 1556
|
|
Up: NOTE 203 - PARALLELIZATION OF AIPS++
Previous: Categorization of User Specifications into Embarrassingly Parallel, Fine-Grain,
Subsections
- Glish changes. Changes to Glish internals to include
machine-specific system calls required to execute a process on a
particular processor must be carried out.
- File access and locking. AIPS++ must have the capability
to lock files and control file access when separate processors are
working on separate parts of data in the EP case.
- Selection of f77 and HPF code. The AIPS++ installation
system must be able to select between f77 and HPF Fortran subroutines,
depending on whether the site has an HPF compiler.
- User input of number of processors. The user interface
must have the ability for the user to indicate how many processors
(ultimately what machine the processors are on) to use for a task
execution.
- Distribute Parallelization Plan to AIPS++ Programmers.
- Implement calls to existing tuned HPF libraries.
- Rewrite existing Glish internals to allow machine calls to
select processor numbers.
- Parallelize simple spectral line imaging and deconvolution
in Glish.
- Import pulsar de-dispersion software, parallelize at fine
grain.
- Parallelize sorting and gridding at fine grain.
- Parallelize antenna-based calibration and self-calibration.
- Investigate parallel solutions to linear problems with
no-linear constraints (similar to Briggs NNLS).
Up: NOTE 203 - PARALLELIZATION OF AIPS++
Previous: Categorization of User Specifications into Embarrassingly Parallel, Fine-Grain,
  Contents
Please send questions or comments about AIPS++ to aips2-request@nrao.edu.
Copyright © 1995-2000 Associated Universities Inc.,
Washington, D.C.
Return to AIPS++ Home Page
2006-10-15