Getting Started

Documentation

Glish

Learn More

Programming

Contact Us

Version 1.9 Build 1367

News	FAQ
Search	Home

Reccomendations

My main reccomendations are as follows:

1.

Decide on a policy for application memory use. Do we want to require that ``core'' applications use as little memory as possible, i.e., trade of programmer convenience for allowing larger problems to be run on smaller machines. While it seems that Image access is the driver, we should attempt to think of applications that require, for example, access to a large number of visibilities at the same time.

2.

That AIPS++ adopt the ``Standard $5,000 machine'' as its goal.

3.

That an individual be appointed to track progress and extant ``Dangers,'' and to report on progress and prospects on a monthly timescale until running on the standard machine is a reality.

4.

That developers routinely run new applications on standard sized problems on standard machines⁵, and immediately report problems, performance and memory related particularly.

5.

Monitor the probable size of an end-user installation.

6.

Occasionally read standard continuum (particularly) and spectral line data sets into AIPS++ and compare the resultant visibility data-set sizes and size of the resultant images with those in other packages. Do this for datasets with and without polarization measurements.

7.

Plan to modify AipsView so that it does I/O on the underlying image data rather than requiring it to (logically) be entirely within memory.

8.

Investigate weakening Glish's promotion to double (and dcomplex). Investigate a way for users to discover which variables are using up all their space.

9.

Investigate Glish's executable size -- is the problem static functions or something else.

10.

Investigate AIPS++ library size, particularly trying to factor out common code in template instantiations.

11.

Complete the implementation of shared libraries for ``common'' code.