random seed <value>

will do that. I'm not sure it's going to have any effect on the ordering of output parasitics, if that's what you're looking for.

You have read my mind. I'm going to test it

My suspicion is that solving that particular problem is going to require buffering all the output and then ordering it with

qsort()

before sending it to the output.

Sadly, using

random seed 0

doesn't avoid the re-ordering of parasitics in the pex extraction, as you said. I think that if the hashing function relies on a seed that can be fixed, all the parasitics are going to be hashed to a specific value and sorting them may not be required.

It's not the hashing function per se. It's that the parasitic capacitances are dropped into a hash table, and the hash key is the address of the structure, so the position in the hash table is actually determined by the memory management system, making it unpredictable in spite of not involving any random number generator.