llListStatistics is a great function but as far as I'm aware SL doesn't employ any form of SIMD in LSL implicitly let alone explicitly as is being asked with this request, as such it is in itself a good example of how this could provide an improvement. If you have a list of 100 or so values, llListStatistics would (when compiled) call up roughly 100 op codes for adds where as with SIMD, it'd create far fewer op codes and the op codes it does call will be doing the same amount of work as llListStatistics but much of it simoultenously.
Had not considered crypto in SL either. Anyone trying to create some form of crypto currency inside of second life I think would likely be violating some form of TOS rule. 🤷♂️
This function could benefit cryptographic communications scripts by speeding up the cryptographic hashing with bitwise XOR but other than that I don't see any value social nor monetary in anything "crypto" related in SL.
Would be interesting for SL to add proper 2d (or even 3d) arrays but I think they're pretty set on sticking with the data types we have.
It would not benefit that one. The proposed function is for speeding up bulk math/logic operations by doing the same operation on 2, 3, or 4 equations at a time.
you would want to assign the output of llSIMD to a list variable first (either appending or prepending it) and then from there you could move it to individual variables if the list variable is not preferred.
I do somewhat agree with your second point that, yes, an implicit SIMD would be nice but unless they where very clear about how it's implemented it'd be hard for us to know exactly how to implement it. Giving an explicit method of implementing it in this proposed method makes it much easier to know exactly when you are benefiting
They likley do optimize basic operations to some degree. However I could not find anything about SIMD. I suppose it's certainly possible they do, but I could not find any documentation about it anywhere.