Skip to content

Instantly share code, notes, and snippets.

@heikoheiko
Last active August 29, 2015 14:10
Show Gist options
  • Save heikoheiko/d76c20a6f83daceb8f56 to your computer and use it in GitHub Desktop.
Save heikoheiko/d76c20a6f83daceb8f56 to your computer and use it in GitHub Desktop.
#!/usr/bin/env python
# -*- coding: utf-8 -*-
"""
Requirements:
- I/O bound: cycles spent on I/O ≫ cycles spent in cpu
- no sharding: impossible to implement data locality strategy
- easy verification
Thoughts:
Efficient implementations will not switch context (threading) when waiting for data.
But they would leverage all fill buffers and have concurrent memory accesses.
It can be assumed, that code can be written in a way to calculate N (<10)
nonces in parallel (on a single core).
So, after all maybe memory bandwidth rather than latency is the actual bottleneck.
Can this be solved in a way that aligns with hashing nonces and allows
for a quick verification? Probably not.
Loop unrolling:
Initially proposed dagger sets offer data locality which allows to scale the algo
on multiple cores/l2chaches. 320MB / 40sets = 8MB (< L2 cache)
A solution is to make accessed mem location depended on the value of the
previous access.
Partitial Memory:
If a users only keeps e.g. one third of each DAG in memory (i.e. to
have in L3 cache), he still can answer ~0.5**k of accesses by substituting
them through previous node lookups.
This can be mitigated by
a) making each node deterministically depend on the value of at
least one close high memory node. Optionally for quick validation, select
the 2nd dependency for the lower (cached) memory. see produce_dag_k2dr
b) for DAG creation, using a hashing function which needs more cycles
than multiple memory lookups would - even for GPUs/FPGAs/ASICs.
"""
try:
shathree = __import__('sha3')
except:
shathree = __import__('python_sha3')
import time
def sha3(x):
return decode_int(shathree.sha3_256(x).digest()) #
def decode_int(s):
o = 0
for i in range(len(s)):
o = o * 256 + ord(s[i])
return o
def encode_int(x):
o = ''
for _ in range(64):
o = chr(x % 256) + o
x //= 256
return o
def get_daggerset(params, seedset):
return [produce_dag(params, i) for i in seedset]
def update_daggerset(params, daggerset, seedset, seed):
idx = decode_int(seed) % len(daggerset)
seedset[idx] = seed
daggerset[idx] = produce_dag(params, seed)
P = (2**256 - 4294968273)**2
def produce_dag(params, seed):
k, w, d = params.k, params.w, params.d
o = [sha3(seed)**2]
init = o[0]
picker = 1
for i in range(1, params.dag_size):
x = 0
picker = (picker * init) % P
#assert picker == pow(init, i, P)
curpicker = picker
for j in range(k): # can be flattend if params are known
pos = curpicker % i
x |= o[pos]
curpicker >>= 10
o.append(pow(x, w, P)) # use any "hash function" here
return o
def quick_calc(params, seed, pos, known={}):
init = sha3(seed)**2
k, w, d = params.k, params.w, params.d
known[0] = init
def calc(i):
if i not in known:
curpicker = pow(init, i, P)
x = 0
for j in range(k):
pos = curpicker % i
x |= calc(pos)
curpicker >>= 10
known[i] = pow(x, w, P)
return known[i]
o = calc(pos)
return o
def produce_dag_k2dr(params, seed):
"""
# k=2 and dependency ranges d [:i/d], [-i/d:]
Idea is to prevent partitial memory availability in
which a significant part of the higher mem acesses
can be substituted by two low mem accesses, plus some calc.
"""
w, d = params.w, params.d
o = [sha3(seed)**2]
init = o[0]
picker = 1
for i in range(1, params.dag_size):
x = 0
picker = (picker * init) % P
curpicker = picker
# higher end
f = i/d + 1
pos = i - f + curpicker % f
x |= o[pos]
curpicker >>= 10
# lower end
pos = f - curpicker % f - 1
x |= o[pos]
o.append(pow(x, w, P)) # use any "hash function" here
return o
def quick_calc_k2dr(params, seed, pos, known={}):
# k=2 and dependency ranges d [:i/d], [-i/d:]
init = sha3(seed) ** 2
k, w, d = params.k, params.w, params.d
known[0] = init
def calc(i):
if i not in known:
curpicker = pow(init, i, P)
x = 0
# higher end
f = i/d + 1
pos = i - f + curpicker % f
x |= calc(pos)
curpicker >>= 10
# lower end
pos = f - curpicker % f - 1
x |= calc(pos)
known[i] = pow(x, w, P)
return known[i]
o = calc(pos)
return o
produce_dag = produce_dag_k2dr
quick_calc = quick_calc_k2dr
def hashimoto(daggerset, lookups, header, nonce):
"""
Requirements:
- I/O bound: cycles spent on I/O ≫ cycles spent in cpu
- no sharding: impossible to implement data locality strategy
# I/O bound:
e.g. lookups = 16
sha3: 12 * 32 ~384 cycles
lookups: 16 * 160 ~2560 cycles # if zero cache
loop: 16 * 3 ~48 cycles
I/O / cpu = 2560/432 = ~ 6/1
# no sharding
lookups depend on previous lookup results
impossible to route computation/lookups based on the initial sha3
"""
num_dags = len(daggerset)
dag_size = len(daggerset[0])
mix = sha3(header + encode_int(nonce)) ** 2
# loop, that can not be unrolled
# dag and dag[pos] depended on previous lookup
for i in range(lookups):
dag = daggerset[mix % num_dags] # modulo
pos = mix % dag_size # modulo
mix ^= dag[pos] # xor
return mix
def light_hashimoto(params, seedset, header, nonce):
lookups = params.lookups
dag_size = params.dag_size
known = dict((s, {}) for s in seedset) # cache results for each dag
mix = sha3(header + encode_int(nonce)) ** 2
for i in range(lookups):
seed = seedset[mix % len(seedset)]
pos = mix % dag_size
mix ^= quick_calc(params, seed, pos, known[seed])
num_accesses = sum(len(known[s]) for s in seedset)
print 'Calculated %d lookups with %d accesses' % (lookups, num_accesses)
return mix
def light_verify(params, seedset, header, nonce):
return light_hashimoto(params, seedset, header, nonce) \
<= 2**512 / params.diff
def mine(daggerset, params, header, nonce=0):
orignonce = nonce
origtime = time.time()
while 1:
h = hashimoto(daggerset, params.lookups, header, nonce)
if h <= 2**512 / params.diff:
noncediff = nonce - orignonce
timediff = time.time() - origtime
print 'Found nonce: %d, tested %d nonces in %.2f seconds (%d per sec)' % \
(nonce, noncediff, timediff, noncediff / timediff)
return nonce
nonce += 1
class params(object):
"""
=== tuning ===
memory: memory requirements ≫ L2/L3/L4 cache sizes
lookups: hashes_per_sec(lookups=0) ≫ hashes_per_sec(lookups_mem_hard)
k: ?
d: higher values enfore memory availability but require more quick_calcs
numdags: so that a dag can be updated in reasonable time
"""
memory = 512 * 1024**2 # memory usage
numdags = 128 # number of dags
dag_size = memory /numdags / 64 # num 64byte values per dag
lookups = 512 # memory lookups per hash
diff = 2**14 # higher is harder
k = 2 # num dependecies of each dag value
d = 8 # max distance of first dependency (1/d=fraction of size)
w = 2
if __name__ == '__main__':
print dict((k,v) for k,v in params.__dict__.items() if isinstance(v,int))
# odds of a partitial storage attack
missing_mem = 0.01
P_partitial_mem_success = (1-missing_mem) ** params.lookups
print 'P success per hash with %d%% mem missing: %d%%' %(missing_mem*100, P_partitial_mem_success*100)
# which actually only results in a slower mining, as more hashes must be tried
slowdown = 1/ P_partitial_mem_success
print 'x%.1f speedup required to offset %d%% missing mem' % (slowdown, missing_mem*100)
# create set of DAGs
st = time.time()
seedset = [str(i) for i in range(params.numdags)]
daggerset = get_daggerset(params, seedset)
print 'daggerset with %d dags' % len(daggerset), 'size:', 64*params.dag_size*params.numdags / 1024**2 , 'MB'
print 'creation took %.2fs' % (time.time() - st)
# update DAG
st = time.time()
update_daggerset(params, daggerset, seedset, seed='new')
print 'updating 1 dag took %.2fs' % (time.time() - st)
# Mine
for i in range(10):
header = 'test%d' % i
print '\nmining', header
nonce = mine(daggerset, params, header)
# verify
st = time.time()
assert light_verify(params, seedset, header, nonce)
print 'verification took %.2fs' % (time.time() - st)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment