its quite possible you have done something similar already:
this is the diehard battery running on rnd.cpp, compared to mt and isaac (not broken so far).
https://stash.reaper.fm/oldsb/940657/...dl_rng_sha.log
https://stash.reaper.fm/oldsb/938662/diehard_mtrand.log
https://stash.reaper.fm/oldsb/940656/diehard_isaac.log
a total of 3e+5 32bit values are used for the above results.
from what i've seen poor crypto-prngs might perform worse in the higher bit ranges of the OPSO, OQSO and DNA tests.
and of course these results give absolutely no guaranties...
for the fun of it you can try putting a small prize tag on a number sequence and see if someone can brake it (i.e. find initial state, or seed etc..).
---