This analysis dealt with determining a definitive contents list for the numerical library.
Investigation of the known useful algorithms began, in particular the iterative solution of large sparse linear systems, Fast Fourier Transforms and Fast Poisson Solvers. Documentation relating to specific, relevant items of software such as KIVA III, MIKE-3, CUBE, etc. were studied.
Contact between different partners have been established to ensure some consistency in the methods used to evaluate the numerical software requirements of the end-user applications. Furthermore, a number of meetings took place to discuss parallelism in the end-user applications, the potential for library software and the form of reporting for associated deliverables.
The most important activities that took place in this period were the visits to the end-user sites by the academic partners. These visits provided the opportunity to have direct discussions between technical staff that brought out important details of the end-user problems that were not immediately obvious from summaries of the problems given by the end-users.
Preliminary reports from the analyzes began to arrive towards the end of March but even at this early stage certain trends were evident.
The most important area of numerical software that was common to all of the application codes was solvers for sparse systems of equations.
Other areas that emerged were:
It was interesting to note that during the course of this reporting period, it was recognized that CERFACS were also going to be an end user. The spectral portraits module of PRECISE can make use of a solver for large sparse complex non-Hermitian matrices and routines for computing selected eigenvalues of large sparse Hermitian matrices.
This task was re-activated following the reviewers' report in the first review. Work was carried out to bring in new applications to the project. In particular, Thomson undertook the examination of one further applications which involves computation of the discrete Fourier Transform of a complex bi-variate sequence.