Check-point file

When an adf calculation terminates abnormally - not controlled by the program itself, for instance after a core dump due to some bug - there will usually be a file TAPE13, which serves as a checkpoint file. tape13 can be used to restart the calculation at a point not too far before the fatal condition occurred. It contains only data for the restart, but none of the special analysis data on TAPE21 that would be useful for analysis, to serve as fragment file, etc.

TAPE13 is upgraded during the calculation but discarded upon normal termination, namely when all relevant information has been saved on TAPE21. At that point all info that would have been on TAPE13 is present on TAPE21. If you wish to keep tape13 anyway - for instance because you plan a restart after normal termination and don't intend to keep the substantially bigger TAPE21 - you must use the save key.

Upon normal (i.e. program-controlled) termination of a calculation, the TAPE21 result file can be used for restart purposes. When a crash occurs, however, chances are that TAPE21 has not correctly been closed and that its data structure is inconsistent: during the calculation large portions of TAPE21 are kept in memory rather than on file, and only at the point of final termination, all data is flushed to file.