Dynamic over-allocation of work space

DFSORT can dynamically over-allocate the work space even when you specify the number of records under the following circumstances:

When you delete a significant number of records using:
- An INCLUDE or OMIT statement, or the SKIPREC option. Use of these statements and options does not force DFSORT to use a SIZE=En or FILSZ=En specification. DFSORT ignores the En value unless it cannot compute the input file size.
- One or more partitioned data set members as input. DFSORT uses the size of the entire partitioned data set rather than the size of the member in its calculations. DFSORT ignores any SIZE=En or FILSZ=En value unless it cannot determine the input file size itself.
You can avoid over-allocation in these cases by specifying SIZE=Un or FILSZ=Un.
When the average record length of variable-length records is substantially shorter than one-half of the maximum record length. If DFSORT uses your exact or estimated number of records, it uses one-half of the maximum record length to determine the file size. You can avoid over-allocation in this case by specifying AVGRLEN=n.

Dynamic over-allocation of work space can occur when you do not specify the number of records (for example, with small input data sets on tape), or even when you do (for example, when a significant number of records is deleted). In these cases, you might prefer to use JCL allocation of work data sets to control the amount of space allocated. However, there are drawbacks to doing so, as previously explained. If DYNAUTO=IGNWKDD is used, remember to specify run-time option USEWKDD when you want to use JCL allocation of work data sets.