Need alleviation from the cost of dynamic linkage with active loadsets

When a program calls another program and the target program is in an active loadset, CCENBK routine CCEHOLDX is executed to chase the PAT chain and locate the correct version of the target program. This is the basic mechanism that allows dynamic linkage to work and allows programs to be loaded and activated in a running TPF system.

The CCEHOLDX overhead is not a "once per ECB" cost. It is incurred every time a function in the target program is called from another program. Consequently, as a program base is modernized to utilize more and more OO designs, the number of "inter-program" calls can increase dramatically and this has a direct impact on the associated increase in CPU of running with active loadsets.

Judicious bundling of program code to minimize "inter-program" calls is one strategy for mitigating those costs, but as business logic becomes increasingly complex it is simply not feasible to split SOs in ways that a minimize inter-program calls to an acceptable level. Furthermore, development contention drastically increases for extremely large SOs, espeically when the compile units they contain are updated frequently.

In order to support continued modernization of business logic on the TPF platform by extending OO designs, we need relief from the CPU cost of the current design.