Hello to everyone,
I just moved from FAST 7.00 to FAST 7.01.00a. As I understand from the changelog there are no modifications to the differential equations solved by FAST. Nevertheless FAST 7.01 needs 40% more computation time than the previous version. Both versions of FAST I compiled with IVF 11, using identical compiler parameters (i.e. optimization settings, /O3). For the comparison I used identical wind turbine models with identical outputs. So none of the new outputs are used by me.
Can anyone confirm, that FAST 7.01 is slower than FAST 7.00?
Are the new outputs calculated internally, independent if the outputs are requested by the user? Can that be an explanation for a 40% increase in computation time?
Does anyone have suggestion what could be done to make FAST 7.01 as fast as its previous version?
A 40% increase seems a lot higher than the tests I ran (I don’t recall the exact numbers right now). However, the outputs you choose can have an impact.
In FAST 7.01.00a-bjj, I modified the algorithms that check if a particular output listed in the input file is valid. With the old method, the IVF 11 compiler usually complained of a loss-of-optimization error. The old method went down a list and compared each value with the requested parameter name; the new method is a binary search. If you choose output parameters that were near the end of the list, the new algorithm will be significantly faster. If the parameters were at the beginning of the list, the new algorhthim will probably be slower. On the CertTests, I observed the total performance to range from unchanged to significantly improved with this modification.
The new outputs are calculated internally for each tower and blade node that have strain gages for output–even if you don’t request the specific parameter for output. However, if NTwGages and NBlGages are both 0, the new outputs will not be calculated.