Hundreds of thousands of lines of Ada code

The Hawk Mission Computer Operational Flight Program (OFP) software comprises hundreds of thousands of lines of Ada source code, executing on a PowerPC (MPC7410) as 25 separate software partitions.

The key objective of the project is to reduce the overall execution time of the OFP Software by 10%. Phase 1 of the project focused on 4 partitions that make up approximately 25% of the schedule. The RapiTime worstcase execution time analysis tool set was used to identify opportunities for reducing the overall execution time of code in these partitions.

Worst-case hotspots

Unlike conventional code profiling techniques, RapiTime identifies worst-case hotspots from the point of view of worst-case execution time. That is the Ada sub-programs and lines of source code that contribute most to the overall worstcase execution time. Conventional profiling techniques identify the lines of code that execute the most on average, which is very different.

Execution time contribution

By combining static analysis of the program structure with timing trace information generated via extensive testing, RapiTime was able to provide information about the contribution of every subprogram to the overall worst-case execution time thus identifying candidate sub-programs for optimisation.

The best candidates for optimisation were inspected. This involved studying both the Ada source code and in some cases the object code produced by the compiler. A variety of code constructs were found to be contributing heavily to the worst-case execution time. Where possible, these constructs were optimised.

Code optimisations

Optimisations included: removing code that created redundant copies of large data structures; re-writing bit unpacking code, enabling the compiler to produce much more efficient object code (in a sub-program that was called over 700 times on the worst-case path); and re-writing case statements using look-up tables (called over 450 times on the worst-case path).

23% reduction in worst-case execution time

Overall, RapiTime accurately identified worst-case hotspots enabling a 23% reduction in the worst-case execution time of the analysed partitions. This reduction was achieved via a targeted and highly focused optimisation effort. This minimised the software development and validation/verification effort required whilst ensuring that the improvements had the maximum possible impact.

By examining 1.2% of the code

The sub-programs that needed to be examined to achieve this reduction represented just over 1% of the code in the selected partitions.

The reductions in worst-case execution time achieved in phase 1 mean that the project is well on track to meet its objectives. Successful conclusion of the project will enable significant extra functionality to be added to the Mission Computer without the need for an extremely costly hardware upgrade.

"We are delighted with the schedule reductions achieved in phase 1" said Dean Armstrong, Software Engineering Manager at BAE Systems Brough. "The integration team are now looking at how RapiTime can be used as an integral part of the software development process for future Hawk developments".