I compiled SC from source on two different Linux computers.
At the beginning of a script, I customize the random generators by experimenting with various seed-numbers in RandSeed.ir(1, xxx); until LFNoise and the other random generators produce outputs that I like.
Unfortunately, when I run the same script on the second computer, LFNoise produces completely different outputs.
Is there a way to ensure that RandSeed remains reproducibly consistent across different installations?
Some time ago, there have been discussions about differences related to implementation-dependent floating point results (sure, random number generators are based on integer math but at some point we’re back in the float sphere server-side). So, maybe, you can never expect a 100% positive answer to that question. Other sensitive calculations in the signal chain could also cause differences for the same reason (filters!).
Furthermore, I’d recommend to take the LFDNoise variants for the reasons described in the help files.
Unfortunately, the remarks are only in the LFDnoise help files (footnote). Granted, if you already like the variants that LFNoise variants produce, that doesn’t help. It could be, though, that LFDNoise variants are less sensitive to those floating point issues, maybe worth a try. Also, as the note states, there’s a samplerate dependency, so check if your linux machines are using the same samplerate.