Discussion Overview
The discussion revolves around finding a Fortran implementation of the Lanczos algorithm suitable for large sparse Hermitian matrices. Participants explore various resources, libraries, and considerations regarding the quality and usability of existing implementations.
Discussion Character
- Exploratory, Technical explanation, Debate/contested
Main Points Raised
- One participant requests a user-friendly Fortran77 source deck for the Lanczos algorithm, expressing difficulty with an existing implementation by Jane Cullum and Ralph Willowghbi.
- Another participant suggests trying the LINPACK library as a potential resource.
- A third participant provides a link to a collection of Lanczos resources on Netlib, indicating it may contain useful information.
- Concerns are raised about the age of the referenced book from 1985, with a participant cautioning that the algorithms may not meet modern standards, especially given the limited size of the largest test problem reported.
- Additionally, a participant recommends ARPACK as a more contemporary alternative for Lanczos eigensolution methods.
- One participant offers a link to a resource that provides in-depth understanding of the Lanczos method for those interested in the theoretical aspects rather than just practical implementation.
Areas of Agreement / Disagreement
Participants do not reach a consensus on a specific implementation or library. Multiple competing views and resources are presented, indicating a lack of agreement on the best approach.
Contextual Notes
There are concerns regarding the quality and applicability of older algorithms, as well as the definitions of "large" eigenproblems in the context of modern computational needs.
Who May Find This Useful
Researchers and practitioners looking for Fortran implementations of the Lanczos algorithm, particularly in the context of large sparse Hermitian matrices, as well as those interested in the theoretical underpinnings of the method.