Two compression methods are available: SCALE or DELTA (see Parameter METHOD).
Further compression is achieved by replacing runs of equal input values by a single occurrence of the value with a corresponding repetition count.
It should be noted that the degree of compression achieved is dependent on the nature of the data, and it is possible for a compressed array to occupy more space than the uncompressed array. The mean compression factor actually achieved is displayed (the ratio of the supplied NDF size to the compressed NDF size).
It is possible to delta compress an NDF that has already been scale compressed. This provides a means of further compressing floating-point arrays. However, note that the default values supplied for DSCALE, DZERO, VSCALE, and VZERO may not be appropriate as they are chosen to maximise the spread of the scaled integer values in order to minimise the integer truncation error, but delta compression works best on arrays of integers in which the spread of values is small.
If the input NDF is already DELTA compressed, it will be uncompressed and then recompressed using the supplied parameter values.
More details of delta compression can be found in Delta Compressed Array Form.
KAPPA --- Kernel Application Package