If a control section is used by several segments, it is usually desirable to place that control section in the root segment. However, the root segment can get so large that the benefits of overlay are lost. If some of the control sections in the root segment could overlay each other (except for the requirement that all segments in a path must be in storage at the same time), the job might be a candidate for multiple region structure. Multiple region structures can also be used to increase segment loading efficiency: processing can continue in one region while the next path to be executed is being loaded into another region.

With multiple regions, a segment has access to segments that are not in its path. Within each region, the rules for single region overlay programs apply, but the regions are independent of each other. A maximum of four regions can be used.

Figure 1 shows the relationship between the control sections in the sample program and two new control sections: CSH and CSI. The two new control sections are each used by two other control sections in different paths. Placing CSH and CSI in the root segment makes the segment larger than necessary, because CSH and CSI can overlay each other. The two control sections should not be duplicated in two paths, because the binder automatically deletes the second pair and an invalid exclusive reference might then result.

Figure 1. Control sections used by several paths

If the two control sections are placed in another region, however, they can be in virtual storage when needed, regardless of the path being executed in the first region. Figure 2 shows all the control sections in a two-region structure. Either path in region 2 can be in virtual storage regardless of the path being executed in region 1. Segments in region 2 can cause segments in region 1 to be loaded without being overlaid themselves.

Figure 2. Overlay tree for multiple-region program

The relative origin of a second region is determined by the length of the longest path in the first region (18000 bytes). Region 2, therefore, begins at 0 plus 18000 bytes. The relative origin of a third region would be determined by the length of the longest path in the first region plus the longest path in the second region.

The virtual storage required for the program is determined by adding the lengths of the longest path in each region. In Figure 2, if CSH is 4000 bytes and CSI is 3000 bytes, the storage required is 22000 bytes, plus the storage required by the special overlay tables.

Care should be exercised when choosing multiple regions. There might be some system degradation caused by the overlay supervisor being unable to optimize segment loading when multiple regions are used.