The analysis of an array variables computes four distinct sets for each program region. These sets are used by both the dependence and privatization tests to determine the safety of parallelization. The data-flow sets for a given region R are informally defined as follows:
¯ ¯ Write: portions of arrays possibly written within region R
¯ ¯ Must Write: portions of arrays always written within region R
¯ ¯ Read: portions of arrays possibly read within region R
¯ ¯ Exposed Read: portions of arrays whose reads are possibly upwards exposed to the beginning of RThese sets are together computed as a 4-tuple transfer function using a backwards region-based analysis as described in Section 3. Because just the transfer function itself is needed, the second phase of the region-based framework can be omitted.