new cpuminer uploaded to github https://github.com/gatra/cpuminer-rminerdaccepted: 222/222 (100.00%), 42193804 khash/s (yay!!!) Now I'm debugging something strange: after mining for a while, the miner says it received this packet: [2014-03-18 14:58:25] < {"params": [8], "id": null, "method": "mining.set_difficulty"} setting difficulty to 8, so obviously I didn't find any more shares after this... looks like I'm back to debugging the server... EDIT: indeed.... logs from server: 2014-03-18 13:57:19,177 DEBUG DBInterface # DB_Mysql_Vardiff INIT ... ... 2014-03-18 13:58:22,251 DEBUG DB_Mysql # Setting difficulty for gatra.user to 8
but I thought difficulty was fixed and vardiff was disabled.... it must be disabled for Riecoin to work. Ok, nevermind, my configuration file was wrong. Just make sure you disable VARDIFF in your stratum-mining if you use Riecoin
|
|
|
is this better than the other miners ? It should be slightly better than the open source official miners for ypool because I precalculate inverses. But it's not better than closed miners that are around here and use larger primorials. Will look into that after the pool server code is stable.
|
|
|
new cpuminer uploaded to github https://github.com/gatra/cpuminer-rminerdaccepted: 222/222 (100.00%), 42193804 khash/s (yay!!!) Now I'm debugging something strange: after mining for a while, the miner says it received this packet: [2014-03-18 14:58:25] < {"params": [8], "id": null, "method": "mining.set_difficulty"} setting difficulty to 8, so obviously I didn't find any more shares after this... looks like I'm back to debugging the server... EDIT: indeed.... logs from server: 2014-03-18 13:57:19,177 DEBUG DBInterface # DB_Mysql_Vardiff INIT ... ... 2014-03-18 13:58:22,251 DEBUG DB_Mysql # Setting difficulty for gatra.user to 8
but I thought difficulty was fixed and vardiff was disabled.... it must be disabled for Riecoin to work.
|
|
|
Gatra: were a fair portion of the shares you found on testnet rejected by stratum-mining?
sometimes I get bursts of invalid shares. In all cases I've seen so far, It's the miner and not the pool. I'm debugging now.
|
|
|
gatra: any updates besides the website?
I pushed to github a few minutes ago. I didn't tag it yet, but everything is there on the default branch. It is needed to recompile everything: - the riecoind daemon had a bug in submitblock (until now all miners used getwork instead of submitblock, but stratum-mining uses it) - the stratum-mining also had bugs in the RPC call to submit a block - fixed a bug with 64bit-overflow in the cpuminer that made it hang and test the same range over and over Now blocks are showing in my wallet (tested with testnet). Please test! After this is confirmed to work, it looks like my next urgent task is an open source optimized miner. I understand how to optimize mine, just didn't have the time to do it yet. My cpuminer is probably the fastest of the ones that use the 2310 primorial, but those using larger primorials (dfg, supercomp) are much faster.... Thanks for the patience, gatra and if it works... please donate!
|
|
|
So they try to sieve for 4 chains specifically at the expense of finding blocks?
This means, as of now when you are mining in a pool, others are stealing your block-solving capabilities.
A quick solution: release the most optimized, "cheating", miner for finding shares specifically. If everyone has access to this miner, people will no longer be "cheating". Yes, the difficulty might fall and pools will find less blocks, but it will only be fair to people who can't modify their mining software to cheat.
Yup. And - they could be, depending on what the pool verifies as a valid block. There's an earlier discussion about this, but it's a little more tricky than I think gatra, jh, or I thought it was now that the sieves are getting huge and the diff is up. As an example, I tested three variants of my best miner: - Version (a) is a "cheat" miner: It ONLY looks for 4 chains; - Version (b) is a "50k-fair" miner: It sieves all six spots to primes < 50,000, just like the open source version does, but then it sieves out the first 4 to 1 billion; - Version (c) is a "block-hunter": It sieves all six to a billion. Version (a) and (b) were indistinguishable. Which means there's no profit with a fast miner to having a worse block-to-share ratio than the average person does. Version (b) is what I'd been running myself previously and releasing. But version (c) is what you really want your miners running, because it finds a lot more blocks. But it produces *shares* at only 66% the rate of version (b), even though it finds more than twice as many blocks. I'm now running version (c) because I think that even though it helps the pool more than it helps me, my smaller-slice-of-a-bigger-pie wins out. But it's tricky. What we need to do is determine the right incentive structure to make miners want to run version (c). One answer is that maybe a00k or supercomputer's code doesn't have the same 1/3rd slowdown for all six that mine does. -Dave If it's 66% then it can be solved this way: a block is a 6tuple of primes, <p1, p2, p3, p4, p5, p6> For shares, we also allow the last 2 numbers to be composite instead of primes, like this: <p1, p2, p3, p4, p5, c6> or <p1, p2, p3, p4, c5, c6> so 4 primes and 2 composites make a valid share. If we allowed the following cases to be shares too: <p1, p2, p3, p4, c6, p6> <p1, p2, p3, c4, p5, p6> <p1, p2, p3, c4, c5, p6> <p1, p2, p3, c4, p5, c6> Then versions (a) and (b) would miss lots of these cases, loosing their advantage. The best share-finder would be also the best block-finder. I see only one drawback: currently a miner can stop testing if it finds something like <p1, p2, p3, c4, ??>. After the 4th test you know that you won't find a block or a share. But with this modification, the miner has to do 2 more tests because it could be a valid share of the form <p1, p2, p3, c4, c5, p6>. This means the pooled miner has a small overhead compared to the solo block-finder that stops after the first composite. However this overhead is really mnimal: how many 3ch/s are you currently having per cpu? less than 1. Probably one 3ch every 10 seconds or more. That's how often you''d incurr in this overhead. Currently, my stratum-mining accepts any combination that has at least 4 primes as long as the first one of the 6tuple is prime. I'm testing more carefully and I have to sleep now, but if everything goes ok we'll have full working new versions of the client, miner and stratum server during tomorrow. gatra EDIT: another thing: if you suspect someone is cheating ypool, you should contact jh00. I think he does trial division tests for some divisors on each share, so your (a) miner would be detected.
|
|
|
Finished a lot quicker than I expected. Mostly because I've been working on this the entire day since I got home. Felt productive ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) . Sent the design to gatra. Hope he can upload it and show you the new designed site. Sleeping now. wow! much nicer! thanks upladed here http://riecoin.org/RiecoinOption2/
|
|
|
found a bug in stratum-mining! when a user submits a block, the share is rejected and it's not shown in the pool stats or the wallet ![Sad](https://bitcointalk.org/Smileys/default/sad.gif) sorry to the brave early adopters, I'm working on it
|
|
|
fixed a bug in the miner, version 3.2.1 is in github
But I don't have a win64 box handy, will compile and officially release later
|
|
|
Mining 4 cores on an Intel Core-i5-4570 on ypool using their standard XPT miner. 7.55 shares on Ypool. approx 2 coins per day. Windows 7 64 bit version 8GB RAM 200GB SSD 2TB HDD.
Thanks Dave for your blog article I loved it.
Is it possible to get a plain english version of the POW?
1. Block code created with all transactions in it. 2. Hash taken of this block. 3. Prime sextuplet found such that all six primes add up the hash. 4. The number of leading numbers ignored = the difficulty factor.
Am I on the right track here? I have a love of mathematics and am prepared to learn C++ in order to help with miner optimization. But I don't have a clear grasp of what the POW is doing. Help wanted.
this part here looks more like Math than Code: Let S be the sha256 of the first part of the block header (the nonce is not needed).
In binary, let basep be an 1 concatenated with eight 0's, concatenated with S, concatenated with as many 0's as needed to reach the required difficulty. Let Z be that quantity of 0's.
Find X such that basep + X is the first prime of a prime constellation of size n, where n is hardcoded at 6. The value of X should not be larger than 2^Z
|
|
|
Working on my pool now.. should be ready in less than 10 hours cool! this is my attempt: http://imgur.com/CWcnfIeInvalid shares are from before fixing the bugs, and the RIC/USD price shown is actually from LTC, but I got only valid shares after fixing everything. Obviously my cron jobs are not ok... but one thing I still don't understand is that the shares table has much more shares than shown in the dashboard.... anyway I think it has to be an mpos problem, not a stratum-mining thing
|
|
|
stratum-mining is accepting shares and writing to the DB! my pool still doesn't work (mpos configuration issues) I'll clean up the code and upload everything later, we should have more pools soon!
give me some hours and I'll do the formal Thursday status post, it may come late, but it will still be Thursday.. in at least one time zone... I'll upload to github the stratum-mining, the cpuminer with stratum support, and a new client 0.8.8 (update required only for pool operators)
price still droping! wtf? I'd say it's botnets dumping...
|
|
|
Anyone think we need a redesign for Riecoin logo? Its a bit 2D in my opinion and doesn't really stand out from the oceans of cryptocurrencies we have right now.
If there is some bounty, i can make a lot more 3D design ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) 50 RIC on my end for a redesign of logo. I like our logo, but if the majority speaks I'll agree to change it. Please consider that 3D is not always better than 2D! well... the Riemann zeta function is actually 4D... (C->C ie 2D->2D) but can be represented in two 3D graphics... Also consider that besides the logo I'll need the splash screen for the QT wallet, the icons in many sizes, the drawing for the left of the "about riecoin" window, the vertical banner for the windows installer, and the horizontal banner like this one: http://www.riecoin.org/logo_riecoin_png.png
|
|
|
cool work! nice chart... One thing to have in mind: the nonce (or offset) can be up to 256 bits, however it also has to fit in the trailing zeros of the base number (the 1 + 8 zeros + the hash + x zeros to fill for the difficulty). Without this rule, 2 blocks could share the same sextuplet. This means that for difficulty 304, the max valid nonce was 39 bits long. After difficulty 521 this didn't matter anymore, but the size of the nonce was limited at the beginning (this is why supercomputing would probably loose his bet of mining the first 500 blocks in 5 mins - also, there is a retarget at 288 which is special: the first retargets don't follow the 4x rule so difficulty could jump up fast) I never used more than 64 bits for the nonce and 2310 for the primorial (which is the cpuminer I published).
|
|
|
It'd be nice if the dev. post updates or replies to our post at least once every other day for an active involvement.
Gatra posts replies constantly on all matters concerning the coin. I think he plans to start releasing status updates every Thursday. I honestly believe if he's not responding to posts instantly, it's because he's either sleeping or at work on code. Right now I think he's working on the stratum protocol so it will be easier to start Riecoin pools. Here's also hoping that the pool code can be adapted for p2pool as well ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) yep, I started with the idea of making a p2pool. But stratum-mining seemed easier and the logical step before p2pool. current status: miner finds and reports shares, but stratum mining server receives the share and discards it as invalid (having 0 primes). It's probably some endianness thing or a bug in how the primes are rebuilt server-side. If I can't fix it for tomorrow's status update post then I'll upload it to github in it's current state, maybe someone can debug it. Have to go now... "real life" job calling me...
|
|
|
|