archive snowblossom slack dev
prev 2018-06-26 next
2018-06-26 04:56:08
Fireduck
I made stupid mistakes yesterday
2018-06-26 04:56:17
Fireduck
the address history work doesn't respect reorgs
2018-06-26 04:56:28
Fireduck
and could show things taht are not in the current main chain
2018-06-26 04:56:28
Fireduck
which is no good
2018-06-26 05:29:33
Fireduck
Anyone running my address history code from last night should do a db nuke and recreate
2018-06-26 05:29:42
Fireduck
with addr_index
2018-06-26 05:29:51
Fireduck
if not using addr_index, there is no difference
2018-06-26 05:31:52
Clueless
@Fireduck good thing I missed you have to do addr_index
2018-06-26 05:37:16
Fireduck
hah
2018-06-26 05:37:25
Fireduck
you don't need it unless you have an explorer attached
2018-06-26 18:45:50
Shoots
Have you guys started working on a BTC style RPC for exchanges to use?
2018-06-26 19:04:10
Rotonen
do exchanges sponsor that sort of work?
2018-06-26 19:25:27
Fireduck
@Shoots not yet. My hope is to get in touch with an exchange and get a rough set of requirements to work towards
2018-06-26 19:25:32
Fireduck
rather than guessing what they want
2018-06-26 19:26:42
Shoots
I'm pretty sure all exchanges look for an RPC that mirrors BTC
2018-06-26 19:27:24
Fireduck
sure, but there are a bunch of questions. Do they want one wallet per customer? Do they want notifications or are they happy to poll when a user logs in?
2018-06-26 19:34:13
Rotonen
@Fireduck if you wanna raise eyebrows and make a splash, don’t do cointopia as the first one
2018-06-26 19:35:10
Fireduck
I am aware of that terrible json rpc protocol
2018-06-26 19:35:49
Shoots
terrible maybe, but thats how most if not all exchanges handle RPC
2018-06-26 19:36:06
Fireduck
sure, and it has about 100 calls. It would be good to know which ones anyone cares about.
2018-06-26 19:36:47
Fireduck
@Rotonen why is cointopia some sort of list anything that moves exchange?
2018-06-26 19:38:29
Shoots
These guys seem like they're pretty decent for getting back to dev's https://crex24.com/
2018-06-26 19:40:22
Fireduck
Are they going to be irritated when I answer "What listing fee do you propose to pay?" with "Nothing" ?
2018-06-26 19:40:36
Fireduck
Or is that a gate question to eliminate terrible scams
2018-06-26 19:40:38
Fireduck
heh
2018-06-26 19:40:42
Fireduck
doesn't really matter
2018-06-26 19:41:43
Tyler Boone
not sure why they would be irritated.
2018-06-26 19:41:56
Shoots
respond with "We are an honest hard working community that believes projects started with no premine deserve a free listing."
2018-06-26 19:43:09
Shoots
DEFT just got listed on that exchange and it actually seems decent. The guy they were dealing with seemed to be responding very fast and they actually set a trading launch date that they hit.
2018-06-26 19:44:25
Fireduck
nice
2018-06-26 19:45:18
Fireduck
alright, I'm all over it
2018-06-26 21:56:51
Rotonen
@Fireduck 'not even on cointopia yet' is a sign of freshness of new currencies
2018-06-26 22:19:04
Rotonen
@Fireduck `INFO: Mining rate: 0.000/sec - at this rate ∞ hours per block` is common when not having many miners, how does it decide that? also claims being stalled
2018-06-26 22:19:28
Fireduck
I should fix that
2018-06-26 22:19:44
Rotonen
``` INFO: Mining rate: 5-min: 1.007M/s, 15-min: 997.312K/s, hour: 808.569K/s Jun 26, 2018 10:18:53 PM snowblossom.miner.MrPlow printStats INFO: Mining rate: 0.000/sec - at this rate ∞ hours per block Jun 26, 2018 10:18:53 PM snowblossom.miner.MrPlow printStats INFO: we seem to be stalled, reconnecting to node ```
2018-06-26 22:19:48
Fireduck
It is assuming no shares in 20 seconds is a problem
2018-06-26 22:20:33
Rotonen
well, especially with the difficulty ramp up creep issues, that'll jam things up a bit
2018-06-26 22:20:38
Rotonen
what's the cost of it hitting that, actually?
2018-06-26 22:27:54
Fireduck
None
2018-06-26 22:28:33
Rotonen
good, just from time to time wondering if there is actually something wrong with the pool as relatively many people try it a bit and leave
2018-06-26 22:29:07
Rotonen
grepping the logs i've had 100+ addresses on within 24h, but like 99% of the shares are mine
2018-06-26 22:29:36
Rotonen
i'll go with 'most people are confused'
2018-06-26 22:32:11
Tyler Boone
what pool do you run?
2018-06-26 22:33:01
Tyler Boone
snowplough?
2018-06-26 22:33:21
Rotonen
yeah, snowplough is mine
2018-06-26 22:33:32
Tyler Boone
I mine on that pool
2018-06-26 22:33:52
Tyler Boone
but it's my gaming machine, so I kill the miner often
2018-06-26 22:34:06
Tyler Boone
by "my gaming machine" I mean "my only computer"
2018-06-26 22:34:38
Rotonen
i guess you're the frequent 100k range miner then
2018-06-26 22:35:07
Tyler Boone
yup
2018-06-26 22:35:40
Rotonen
you do about the same as i do, throw spare capacity at it
2018-06-26 22:36:07
Rotonen
though mine bursts to some 30M range if everything is idle (but this essentially never happens, though, as my infra is mostly actually doing stuff)
2018-06-26 22:37:27
Rotonen
the one i have constantly there just to show there is any mining going on, is actually testing over different combinations of jvm and miner parametres slowly
2018-06-26 22:37:44
Tyler Boone
haha, nice
2018-06-26 22:38:29
Rotonen
so far nothing is out of the ordinary very much
2018-06-26 22:39:18
Rotonen
but i actually reach the same terminal mining speed without memfield as with memfield, it just takes like 2 days
2018-06-26 22:40:15
Rotonen
so memfield and hybrid are actually only mitigate a slow warmup
2018-06-26 22:41:27
Tyler Boone
the max depth parameter will make a difference
2018-06-26 22:41:46
Tyler Boone
if you can't fit the entire field into memory
2018-06-26 22:43:01
Rotonen
what's that actually controlling and is that on a branch?
2018-06-26 22:43:28
Tyler Boone
that causes the miner to drop a piece of work on the floor if it has to go to disk early in the hash chain
2018-06-26 22:43:37
Tyler Boone
each POW has 6 hashes (as you know)
2018-06-26 22:44:13
Tyler Boone
so if you set it to 2, then if the first or second hash can't be served from memory, it will drop it on the floor
2018-06-26 22:44:42
Rotonen
i think i'd rather let it pull new stuff into the cache - that'll average out over time
2018-06-26 22:45:03
Rotonen
but i suppose if you precache as in hybrid, that'd then pull more hashes, but decrease the collision chance?
2018-06-26 22:45:05
Tyler Boone
this is for miners who can't cache the entire snowfield
2018-06-26 22:45:09
Rotonen
is that actually a good thing?
2018-06-26 22:45:34
Rotonen
yeah, but that'll kink the search distribution, right?
2018-06-26 22:45:46
Tyler Boone
yeah, but who cares?
2018-06-26 22:46:11
Rotonen
i'm not convinced that's actually conductive to block hit rates
2018-06-26 22:46:43
Tyler Boone
when I had 51gb of 64gb in memory it bumped my hashrate from ~110k to ~230k
2018-06-26 22:47:02
Tyler Boone
gotta run
2018-06-26 22:47:03
Tyler Boone
ttyl
2018-06-26 22:47:04
Rotonen
sure, but quality of collision attempts makes that apples to oranges
2018-06-26 22:47:24
Tyler Boone
nah, it's just trading extra CPU for less disk io
2018-06-26 22:47:24
Rotonen
or rather you're trading entropy quality off for density
2018-06-26 22:47:33
Rotonen
rather do zram or something for that
2018-06-26 22:47:49
Rotonen
the miner itself gets bumped into zram first
2018-06-26 22:48:00
Rotonen
and that compresses well, also thread overlaps do
2018-06-26 22:49:39
Rotonen
if you do that across many miners, and you can have offsets for the precache range, and you have enough miners to cover the whole field, that will work
2018-06-26 22:49:59
Rotonen
but that'd require the pools to orchestrate that and the miners to ask for allocation ranges from the pools
2018-06-26 22:50:13
Rotonen
so summa summarum i do see a use for the feature
2018-06-26 22:50:58
Rotonen
that'd actually allow everyone to do ram-only or hybrid, but share value counts get a lot more complicated
2018-06-26 23:04:33
Tyler Boone
the case can be made pretty easily using an argument ad absurdum
2018-06-26 23:05:40
Tyler Boone
imaging you have 1GB of storage with almost unlimited throughput (like 1pb/sec), and all other storage is extremely low throughput (like 10 bytes/sec).
2018-06-26 23:05:50
Tyler Boone
and you have 10,000 cores
2018-06-26 23:06:33
Tyler Boone
it will be more effective to use all those cores to search for a POW which does all 6 hashes within the 1GB of ulta-fast storage
2018-06-26 23:06:47
Tyler Boone
that it would be to use the 10 byte/sec throughput
2018-06-26 23:07:04
Fireduck
Miners are already segregated with separate nonce prefixes
2018-06-26 23:07:43
Fireduck
WTB one tboone machine
2018-06-26 23:13:18
Tyler Boone
BTW, @Rotonen, the feature is in master, but I don't know if it is in a release. you can test it out by using min_depth_to_disk=<x> in your miner's config file.
2018-06-26 23:13:37
Tyler Boone
it's only applicable if using memfield_precache_gb for some amount smaller than the entire snowfield