The program's performance can be defined in terms of the amounts of time (cpu and i/o seconds) and disk space used in a calculation. Also important for the human user is the turn-around time. On multi-user machines cpu-cheap jobs may take a lot of real time to execute due to i/o scheduling.
Therefore it can be a good idea to recompute some items rather than store them on disk. This will increase the amount of cpu time but reduce disk access and it may also improve the turn-around. Another consideration is of course that storage of data on disk may exhaust the available disk space in case of big calculations so that recalculation rather than storage is unavoidable.
DISK {{no}fit} {{no}basis}
instructs adf how to handle the values of the fit functions and basis functions in all integration points: calculate once and store on disk or recompute whenever needed. The (optional) arguments are fit or nofit, and basis or nobasis.
fit and basis tell adf to store the corresponding data on file; the prefix no induces recalculation whenever the data is needed.
Defaults are nofit and nobasis: direct-SCF mode for both features (this can be modified at the installation of adf, see the Installation Manual).
The key DISK has replaced in adf 2.0 the key directSCF in adf 1.x, and extended the applicability of the I/O versus recalculation choice from fit functions-only to basis functions as well.
