A “Future Work” item described in the README_FAST.pdf for version v8.16.00a-bjj is as follows:
“Optimize the code, particularly BeamDyn and ElastoDyn, so that it runs faster. We have put our effort into getting the framework to work and hope to address computational efficiency later. We expect improvements in efficiency as development continues.”
Have users found code performance to be an issue?
Have you a measure of the relative speed of each module for a given solution iteration: Are ElastoDyn and BeamDyn the bottlenecks? Is AeroDyn optimised sufficiently as it stands?
Enabling BeamDyn in FAST v8 causes a dramatic decrease in computational speed due to the more sophisticated computations involved and the smaller time steps required. The computational cost has been an issue for some users. We have used a profiler to analyze the performance of different parts of the FAST source code, but there is certainly more work to be done in this area. BeamDyn, ElastoDyn, and the FAST glue/driver code are areas where we’d like to spend time on code clean-up, algorithm improvement, and code optimization, if we can find the time/resources to do so.