Closed mreinstein closed 7 years ago
seedrandom(0) uses zero as the seed. It's different from seedrandom() which autoseeds.
It would be nice to add more command-line options to run_diharder.sh and bitgen.js to allow testing of other seeds (or autoseeding), and the other PRNGs, through the diharder suite. I have not tested the other algorithms included in the package: they are all included based on tests that have been done by their original authors.
seedrandom(0) uses zero as the seed.
Is there any particular reason for seeding with 0 instead of another specific value?
Why is the dieharder test script seed hardcoded? I would think the test harness might produce a different set of random values on each invocation and provide more results.
Zero is not special; it's just a seed.
It would be fine to add command-line options to run_diharder.sh to run it in different modes, including using any fixed seed you choose or a randomly chosen seed. If you're interested in seeing this, I'd be happy to merge an update that adds this.
ah ok. I think that makes sense; so it's really just a quick treatment showing that for at least one specific seed, this does not fail the dieharder test suite.
Thanks for clarifying!
@davidbau your dieharder test harness is really well written!
One part that I'm a little curious about: https://github.com/davidbau/seedrandom/blob/released/test/bitgen.js#L28
what does
seedrandom(0)
do? Is this autoseeding? Essentially the same asseedrandom()
?I noticed that only ARC4 is tested. Are the others worth running through dieharder, or are they known to fail some of the tests?