Colin Percival comments on Litecoin scrypt


SUBMITTED BY: Guest

DATE: May 3, 2013, 7:34 a.m.

FORMAT: Text only

SIZE: 1.5 kB

HITS: 5199

  1. 10:04 -!- Irssi: Join to #litecoin-dev was synced in 30 secs
  2. 14:26 * cperciva waves
  3. 14:27 < cperciva> aspect_ invited me in here to talk about scrypt
  4. 14:35 < cperciva> here's what I wrote in reply to an email from him
  5. 14:36 < cperciva> The distinction between SHA256 and scrypt is in the cost to compute a single
  6. 14:36 < cperciva> hash; since bitcoin (and I assume litecoin) scales the number of hashes needed
  7. 14:36 < cperciva> to create a block, this balances out the increased computational complexity.
  8. 14:36 < cperciva> What you are left with then is the fact that (for optimal efficiency) scrypt
  9. 14:36 < cperciva> requires more RAM, which makes it harder to put onto an ASIC.
  10. 14:36 < cperciva> What constitutes "fair" is, of course, an entirely political question, not a
  11. 14:36 < cperciva> cryptographic one.
  12. 14:36 < cperciva> If your goal is to have something which works well on GPUs but is hard to put
  13. 14:36 < cperciva> onto an ASIC, your current "small scrypt" approach is probably about right.
  14. 14:36 < cperciva> You won't block ASICs completely of course, but you've reduced their advantage
  15. 14:36 < cperciva> by ~ a factor of 10, which may be enough to keep them away for now at least.
  16. 15:25 <@aspect_> hi! thank's for stopping by!
  17. 15:27 <@pooler> cperciva: as someone who spent many hours trying to optimize scrypt for CPUs, I must say I'm honored to see you here :)
  18. 15:28 < cperciva> pooler: I'm honoured that people spend many hours trying to optimize my code :-)

loading...

comments powered by Disqus