QBAL computes respondent-level weights to match a sample to a target population on the marginal proportions of multiple variables simultaneously, a process often called "Sample Balancing" or "Raking."
QBAL uses the Deming-Stephan iterative proportional fitting (IPF) method with an algorithm that is limited by the number of weighting variable points found in the data, rather than the number of possible cells or of cases balanced. This allows very large samples to be balanced quickly on a large number of variables.
The QUIP specification language allows for infinitely complex variable definitions. Samples can be projected to any population size and input data sets with pre-existing weights may be used, allowing for multi-stage as well as multi-variable weighting.
QBAL User's Guide (pdf).
See also how to use QTAB to adjust weight targets for item non-response.