Arnold carried a companion software archive for John C. Nash's Compact Numerical Methods for Computersfrom the second edition onwards. The archive shipped the algorithms described in the book — small-memory linear algebra, least squares, and function minimisation — in the several programming languages that successive generations of readers used to translate the methods into their own work.
What the support archive carried
- Fortransource for each algorithm keyed to the chapter structure — compiled and tested under standard Unix, DOS, and VMS Fortran compilers of the period.
- Pascal translations of the Fortran routines, provided for teaching contexts where Pascal was the native language.
- BASIC implementations of the most compact algorithms, suitable for the microcomputer teaching environments of the 1980s and 1990s.
- R implementations added in the later revisions, aligned with the R-language numerical libraries the author himself contributes to.
- Errata and commentary on specific algorithms, including guidance on numerical stability and convergence behaviour for the trickier minimisation methods.
Using the software today
Most of the algorithms covered in Compact Numerical Methods for Computers are also available in the R standard numerical libraries and in the optimxpackage (J. C. Nash and colleagues) on CRAN, which packages modern variants of the optimisation routines discussed in the book. For researchers reproducing specific historical computations — or teaching the algorithms with period-appropriate code — the Fortran and Pascal sources remain the most faithful reference.
For historical access to the Arnold-era software archive, contact us with the edition you are working from and the specific algorithm or chapter of interest.