[ANN][RIC] Riecoin: constellations POW *CPU* HARD FORK successful, world record

[northranger79510]

@gatra and @northranger79510:

Was the main technical PoW explanation removed from riecoin.org? I was about to reference it. Think it used to be after "What is new about Riecoin" and before "How is Riecoin different from Primecoin"

You're right, I thought it was there but I couldn't find it on the new site. We need to add this back asap!
In the mean time, you can still find it in the old site here: http://riecoin.biz/

Yikes. Looks like I've goofed up on some part while I was rushing. Will look at everything this weekend.

[1714123984]

1714123984

[northranger79510]

Gatra, if you don't mind, can you give me FTP access so I can just upload it directly instead of emailing you? It would be faster.

[c.figgis]

any pool?

My pool will launch in a few more hours. It is deployed on a high performance cluster in multiple datacenters so it should be quite fast and resistant to attacks.

In order to mine with me you will need any miner that supports the stratum protocol such as the main riecoin miner linked on the first post: https://bitcointalk.org/index.php?topic=446703.msg4912512#msg4912512

Dave (dga) has been so kind as to open source his very fast mining client but it does not support stratum. Please encourage gatra to pull Dave's optimizations into the default miner or someone else to add stratum to Dave's miner.

Right now a single pool controls the Riecoin network with over 90% of the computing power. Riecoin will become much more valuable to all of us once other pools come online to distribute the power.

[surfer43]

So we will don't have an alternative pool beside yPool?

We do have upcpu, but that has the latency issues. And technically we have Infinitypool, although it is not worth it to mine there as the miner isn't optimized.

[northranger79510]

So we will don't have an alternative pool beside yPool?

We do have upcpu, but that has the latency issues. And technically we have Infinitypool, although it is not worth it to mine there as the miner isn't optimized.

Darn, luckily we have technologically gifted community members that can hopefully fix this.

[surfer43]

Cool, so now we are about to get our 5th pool. 

[surfer43]

Riecoin difficulty going to go into the 1400s in 90 minutes or less 

[gatra]

any pool?

My pool will launch in a few more hours. It is deployed on a high performance cluster in multiple datacenters so it should be quite fast and resistant to attacks.

In order to mine with me you will need any miner that supports the stratum protocol such as the main riecoin miner linked on the first post: https://bitcointalk.org/index.php?topic=446703.msg4912512#msg4912512

Dave (dga) has been so kind as to open source his very fast mining client but it does not support stratum. Please encourage gatra to pull Dave's optimizations into the default miner or someone else to add stratum to Dave's miner.

Right now a single pool controls the Riecoin network with over 90% of the computing power. Riecoin will become much more valuable to all of us once other pools come online to distribute the power.

I'm looking at dga's code and trying to understand it in order to decide if it's better to add his optimizations to my miner or to add stratum support to his......

[Supercomputing]

Yes only 1 person is mining. Not many will mine on the pool because the miner for stratum is currently 8 times worse at finding blocks than xptMiner.

Hi mate, I will help you out  this weekend although my play time for now is almost nonexistent. Unless I missed something, it should only take a few minutes to place the mining core in the Stratum based miner. If gatra gets to doing it first, then the better.

It would be fair to say that all of the open and closed source miners are now equal in performance with only one exception, large integer arithmetic using AVX2. So far, phase one has been a trivial exercise in saturating the CPU's data caches (L1- L3). Phase two will be a non-trival exercise using quadratic fields to predict search intervals with higher probability of containing prime sextuplets. Although I am not quite sure of how to bypass the limitations imposed by the double SHA-256 output and the 256-bit nonce. Suffice it to say, it will be a fun and interesting exercise that if successful, we will easily break world records.

[dga]

Yes only 1 person is mining. Not many will mine on the pool because the miner for stratum is currently 8 times worse at finding blocks than xptMiner.

Hi mate, I will help you out  this weekend although my play time for now is almost nonexistent. Unless I missed something, it should only take a few minutes to place the mining core in the Stratum based miner. If gatra gets to doing it first, then the better.

It would be fair to say that all of the open and closed source miners are now equal in performance with only one exception, large integer arithmetic using AVX2. So far, phase one has been a trivial exercise in saturating the CPU's data caches (L1- L3). Phase two will be a non-trival exercise using quadratic fields to predict search intervals with higher probability of containing prime sextuplets. Although I am not quite sure of how to bypass the limitations imposed by the double SHA-256 output and the 256-bit nonce. Suffice it to say, it will be a fun and interesting exercise that if successful, we will easily break world records.

I concur.  I was also amused to see how similar jh's release and mine were, at least from the perspective of the optimizations we did.  Mine has one more - sieve extensions / segmentation - but that's about it. I really like a00k's tricks for reducing memory use by early-discard of candidates, and I plan to steal them for my miner, but this is in the range of a 5%-20% optimization, not anything major.  So - perhaps phase 2.1 will be an exercise in memory reduction as well.  There's a lot of redundant storage between threads in all of the open source miners.  There are some other similar optimizations, so I'm sure we can all get another 40-50% out through pure engineering, but then we're stuck for a while until we get some asymptotic improvements.

[dga]

I've written up a new blog post, part #1 of 2, about the basics of engineering a faster Riecoin miner:

http://da-data.blogspot.com/2014/03/fast-prime-cluster-search-or-building.html

Comments and feedback appreciated.  Python code is all available in the same github that you've already downloaded for the fast miner:

https://github.com/dave-andersen/fastrie

You should be able to play with them at home.

[anonuser777]

What is the largest prime sextuplet that's been found so far? Is there any place where we can get this info? It would be nice to see how much progress you guys are making.

[surfer43]

I've written up a new blog post, part #1 of 2, about the basics of engineering a faster Riecoin miner:

http://da-data.blogspot.com/2014/03/fast-prime-cluster-search-or-building.html

Comments and feedback appreciated.  Python code is all available in the same github that you've already downloaded for the fast miner:

https://github.com/dave-andersen/fastrie

You should be able to play with them at home.

Wow, very informative, thanks.

[Supercomputing]

What is the largest prime sextuplet that's been found so far? Is there any place where we can get this info? It would be nice to see how much progress you guys are making.

+1

p=
39608060997774685965671539735535460088759088396278206106961139\
48212987294861666726212878183363138783650236425145150119535113\
44262563770181101029133114675891844716307578888768933908340995\
97000943339665515654207890738930432071209372898499461739109170\
13020474770223203976146958611730374404979010445683910507639185\
47002764884278868255285018617579167544748700688954291834032754\
95894026569298524525497121270791985004410164938416308475088390\
64963728343467018534414285106714674500471916034692457522770203\
644165770602176536729107420200344087193831031327594293816511127

The size of p is 1856-bit or 559 decimal digits:
{p, p+4, p+6, p+10, p+12, p+16} are all primes

[Supercomputing]

I've written up a new blog post, part #1 of 2, about the basics of engineering a faster Riecoin miner:

http://da-data.blogspot.com/2014/03/fast-prime-cluster-search-or-building.html

Comments and feedback appreciated.  Python code is all available in the same github that you've already downloaded for the fast miner:

https://github.com/dave-andersen/fastrie

You should be able to play with them at home.

Good write-up professor Andersen, I enjoy reading your blog posts.

The Bitcoin proof-of-work function (less general)
SHA256( SHA256(block header) ) < t

[istudy92]

Im running b11 on windows 7, i found 44 shares..yet 41 are invalid, am I doing something wrong?

http://www.cs.cmu.edu/~dga/crypto/ric/

made my batch file and using upcpu pool

[tersagun]

Im running b11 on windows 7, i found 44 shares..yet 41 are invalid, am I doing something wrong?

http://www.cs.cmu.edu/~dga/crypto/ric/

made my batch file and using upcpu pool

Might be because every block is solved by ypool. This is what people refer to "51% attack"
Just come to ypool I would say, at least for now.

[dga]

Im running b11 on windows 7, i found 44 shares..yet 41 are invalid, am I doing something wrong?

http://www.cs.cmu.edu/~dga/crypto/ric/

made my batch file and using upcpu pool

I don't believe that upcpu is compatible with the new ypool definition of a share.  You can probably mine there successfully using the earlier -b9 version, but I'm not going to maintain a version of the miner using the "first four" definition - it's too easy to cheat.  The new share definition is much more robust and helps ensure that pool miners are getting an appropriate return for their mining effort.

  -Dave

[dga]

Im running b11 on windows 7, i found 44 shares..yet 41 are invalid, am I doing something wrong?

http://www.cs.cmu.edu/~dga/crypto/ric/

made my batch file and using upcpu pool

Might be because every block is solved by ypool. This is what people refer to "51% attack"
Just come to ypool I would say, at least for now.

For correctness - this is not a 51% attack.  Ypool's dominance means they could implement a 51% attack in various ways, but they aren't.  Upcpu has relatively few orphaned blocks.  They're just not as fast because they don't have as many miners.

The issue is one of a software incompatibility.  I synchronized my update to xptMiner with ypool's update to its share definition to ensure fair mining.

[istudy92]

Alright great, it worked using ypool instead of upcpu pool.
+1