[Cryptech Tech] First hash via EIM

[Paul Selkirk]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I ported hash_tester.c to the current novena design, because it seemed
the easiest thing to do at this point. Initial tests are satisfying -
there were *no* hashing failures, and performance looks good (without
any tuning):

TC4: Read name and version from SHA-256 core.
This just does 3 4-byte reads.
52276017 iterations in 187.356 seconds (279019.562 iterations/sec)
If we define an EIM operation as one read or write, then this is
837059 operations/sec

TC5: SHA-256 Single block message test as specified by NIST.
This does 16 4-byte writes to the block, then one write to the control
register. It does 2 reads of the status register (first time returns
busy, next time returns done), then it does 8 reads to get the digest,
so a total of 17 writes, 10 reads.
27957959 iterations in 601.631 seconds (46470.300 iterations/sec)
or 1254698 EIM operations/sec

TC6: SHA-256 Double block message test as specified by NIST.
This writes one block, writes init, reads status, reads the
intermediate digest, then repeats for the second block, so exactly
double the number of operations, or 34 writes, 20 reads.
9871382 iterations in 424.815 seconds (23236.894 iterations/sec)
or 1254792 EIM operations/sec

TC7: SHA-256 Huge message test.
This writes 1 block, writes the control register 1000 times, with 2
status reads in between, then reads 1 digest, so 1016 writes, 2008 reads.
309741 iterations in 777.014 seconds (398.630 iterations/sec)
or 1205457 EIM operations/sec.

OTOH, the 403 error on http[s] access seems to be related to my
pushes, somehow. Rob, please whack again, and let's try to get to the
bottom of this.

				paul

On 02/04/2015 12:14 PM, Joachim Strömbergson wrote:
> Aloha!
> 
> Paul Selkirk wrote:
>> I've actually been following your work quite closely for the
>> past couple of days. I'll try to get a stress test running
>> today.
> 
> Great! If you have access to a Novena (I'm on a train) you _should_
> be able to run the single block hash test code I just committed and
> see if you get the correct results printed. (Not self testing just
> yet.) If that works we add self test of the expected vector (in the
> code comment) and just run a gazillion iterations.
> 
> I probably won't be able to do it tonight so if you do it you can 
> hopefully claim to be first to know if our baseline works.
> 
> 
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQEcBAEBAgAGBQJU0nVEAAoJELAqzRn9CEWT4wcH/3InWY/9b8f/gDz9VgdGLFw2
UXi9vcFJ/QolIL9dYAvcwb9g5IqqF+MJMIcQfHEtunQFJU5gWWlM83vIUubch4nZ
EmGn1CIb8i4PAV+VWMxu1wf7v0tCzfpdMmfaTBhnsPaA1Klj3zFgot18jVVhC4gt
ruCj1iKixGOuCN++IhnTPgDW2Nu0mzdysl77U+ntXyGRkk5pnSDLUJ3Mj7+ptuQw
wsMvczoA/zmSB9H/ZV7P+2zEASvUguX4poH4EIDrhu6KoYEk36zyY5bV07a+ipxt
NG3sqzU3u6kIXj8Km6qv5Z24vIbz7wS5n67o+rbOvoJWQFYpHD5jGoqHrInqCUs=
=Wawe
-----END PGP SIGNATURE-----