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

[bsunau7]

(a)  I'm releasing these in part because I want to make an argument for getting some smart people (hi, Gatra, jh, grin) to think a little more about the comparison metric for miners.  It's really tricky to compare miners at this early stage of riecoin, as you'll see if you run these miners head-to-head with xptMiner:  They report many fewer 2ch and 3ch/sec, but should be substantially faster at 4ch and 6ch/sec.  But this means that the only meaningful way to compare them is to run them head-to-head for, say, a day and count shares.

The existing miners also do not have the same throughput at different difficulties.  The ones I released above are tuned for today's current diff of about 1200, with a substantially larger prime sieve than gatra's or jh's (about 16 million primes entering the sieve).  But that means they'd probably be slower for low difficulty.  I can't think of a great way to come up with a difficulty-independent metric for comparing miners.

Perhaps, in the longer term, a benchmark suite of test inputs at different difficulties.

Very interested in this one, mostly so I know how I am tracking against the rest of the mining community (i.e am I wasting my time and should I just use someone else's miner).

The metric I am currently using is "time per 2*32 nonce" @ difficulty.  This is a far from perfect metric but gives me a chance to see better/worse type comparisons for my builds.

For reference... I am running at ~2.3s @ 1100 and ~2.5s @ 1200 on an AWS c1.xlarge instance.  I don't want invest time squeezing the last 20-30% out of my current strategy if I am miles away from current best performance.

So please share!

--
bsunau7

[1713930620]

1713930620

[1713930620]

1713930620

[causevd]

For those who don't follow ypool chat, I've put online some toys for people to play with.  These should be considered very very beta quality and they don't come with any support (but if you see me on ypool, feel free to ask -- I'm "squaggle" there, for lack of creativity in picking a name).

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

They're derived from jh00's miner but with a substantial redesign of the way it handles sieving.  I think it's the fastest public pool miner out there, but I'd love to be proved wrong.  Linux only.  Grab the one that's right for your CPU.  They include a 2.5% dev fee that can't be decreased from the command line.

(a)  I'm releasing these in part because I want to make an argument for getting some smart people (hi, Gatra, jh, grin) to think a little more about the comparison metric for miners.  It's really tricky to compare miners at this early stage of riecoin, as you'll see if you run these miners head-to-head with xptMiner:  They report many fewer 2ch and 3ch/sec, but should be substantially faster at 4ch and 6ch/sec.  But this means that the only meaningful way to compare them is to run them head-to-head for, say, a day and count shares.

The existing miners also do not have the same throughput at different difficulties.  The ones I released above are tuned for today's current diff of about 1200, with a substantially larger prime sieve than gatra's or jh's (about 16 million primes entering the sieve).  But that means they'd probably be slower for low difficulty.  I can't think of a great way to come up with a difficulty-independent metric for comparing miners.

Perhaps, in the longer term, a benchmark suite of test inputs at different difficulties.

(b)  These implement the "fair to 50k" policy I discussed earlier:  They sieve all six candidates up to 50,000, and then apply more aggressive sieving only to the first 4 locations, in order to produce the same 4ch-to-6ch rate as the released xptMiner.

(c) How's your miner, Supercomputing?  The release above is a pretty straightforward derivative of jh's, using the same 2310*n+97 search pattern, as was also adopted by gatra's new optimized miner.  The only changes are in the management of the sieve itself to use the vector units and to be more careful about the cache while not wasting work done in computing offsets relative to z_target.  I'd love to know if you've made a more aggressively different design work well.

  -Dave

Thanks for the great speech anyway I am on ypool almost 24/7 and I see your very often there. Also I appreciate your help that you gave there.
Thanks for that and for the explaining. This helped me out to solve some things that I have remembered as false in my head.

Regards causevd,

[gatra]

As I think Supercomputing mentioned, using larger numbers instead of 2310*n+97, say some number 200 bits long instead of 2310, could go a long way.
Regarding the metric, I proposed "range scanned / s @ diff", but "time per 2*32 nonce @ difficulty" would probably work just as well, just don't forget to adjust for the numbers you are skipping (4 out of 5 in the 2310 case). I agree that it's difficult to compare between different difficulties...



Regarding the roadmap, cool! Very good as a start. I'd add Android wallet next to the Mac wallet, I've been thinking on taking on those 2. Also GPU miner.
Open source stratum-mining is my first priority now, it will get many pools and it's easier to do than p2pool, which should be later on the list. I firmly believe we have the potential to grow big once there are lots of pools online, since that will bring more decentralization and trust.
Also get everyone to accept RIC! and website and video in many languages
But don't disregard what goin2mars says: faucets, giveaways, recognition, TOR, news articles. We got an article on NewScientist! let's get more
Let's complete the article in wikipedia. Also there's a wiki here: http://www.reddit.com/r/RieCoin/wiki/index


it's too late here, need to get some sleep now...

[steban]

As I think Supercomputing mentioned, using larger numbers instead of 2310*n+97, say some number 200 bits long instead of 2310, could go a long way.
Regarding the metric, I proposed "range scanned / s @ diff", but "time per 2*32 nonce @ difficulty" would probably work just as well, just don't forget to adjust for the numbers you are skipping (4 out of 5 in the 2310 case). I agree that it's difficult to compare between different difficulties...



Regarding the roadmap, cool! Very good as a start. I'd add Android wallet next to the Mac wallet, I've been thinking on taking on those 2. Also GPU miner.
Open source stratum-mining is my first priority now, it will get many pools and it's easier to do than p2pool, which should be later on the list. I firmly believe we have the potential to grow big once there are lots of pools online, since that will bring more decentralization and trust.
Also get everyone to accept RIC! and website and video in many languages
But don't disregard what goin2mars says: faucets, giveaways, recognition, TOR, news articles. We got an article on NewScientist! let's get more
Let's complete the article in wikipedia. Also there's a wiki here: http://www.reddit.com/r/RieCoin/wiki/index


it's too late here, need to get some sleep now...

You are doing a great job, and we are here to support you.

[dogaatac]

For those who don't follow ypool chat, I've put online some toys for people to play with.  These should be considered very very beta quality and they don't come with any support (but if you see me on ypool, feel free to ask -- I'm "squaggle" there, for lack of creativity in picking a name).

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

They're derived from jh00's miner but with a substantial redesign of the way it handles sieving.  I think it's the fastest public pool miner out there, but I'd love to be proved wrong.  Linux only.  Grab the one that's right for your CPU.  They include a 2.5% dev fee that can't be decreased from the command line.

(a)  I'm releasing these in part because I want to make an argument for getting some smart people (hi, Gatra, jh, grin) to think a little more about the comparison metric for miners.  It's really tricky to compare miners at this early stage of riecoin, as you'll see if you run these miners head-to-head with xptMiner:  They report many fewer 2ch and 3ch/sec, but should be substantially faster at 4ch and 6ch/sec.  But this means that the only meaningful way to compare them is to run them head-to-head for, say, a day and count shares.

The existing miners also do not have the same throughput at different difficulties.  The ones I released above are tuned for today's current diff of about 1200, with a substantially larger prime sieve than gatra's or jh's (about 16 million primes entering the sieve).  But that means they'd probably be slower for low difficulty.  I can't think of a great way to come up with a difficulty-independent metric for comparing miners.

Perhaps, in the longer term, a benchmark suite of test inputs at different difficulties.

(b)  These implement the "fair to 50k" policy I discussed earlier:  They sieve all six candidates up to 50,000, and then apply more aggressive sieving only to the first 4 locations, in order to produce the same 4ch-to-6ch rate as the released xptMiner.

(c) How's your miner, Supercomputing?  The release above is a pretty straightforward derivative of jh's, using the same 2310*n+97 search pattern, as was also adopted by gatra's new optimized miner.  The only changes are in the management of the sieve itself to use the vector units and to be more careful about the cache while not wasting work done in computing offsets relative to z_target.  I'd love to know if you've made a more aggressively different design work well.

  -Dave

i couldnt worked your miner about 2-3 hours

[Supercomputing]

Yes, I am using the maximum primorial which fits in my sieving interval without exceeding the 256-bit limit.  Also, 11# (2310) works well around 300-bit numbers in size but begins to lose steam as the numbers increase in size.

I probably have the slowest pool miner, but I may very well have the fastest solo miner for now. I implemented my own closed source modular exponentiation library using AVX2 instructions, which gives me about a 30% increase over the latest version of GMP. However, the GMP developers will soon use these new instructions in their next release.

Also, apart from the obvious enhancements of optimizing the usage L1/L2 data caches and eliminating all unnecessary division instructions, I am mining directly on the wallet to eliminate the getwork protocol overhead. We need an efficient protocol similar to XPT (x.pushthrough), to broadcast new block changes to all connected solo miners.

Going forward, I need to benchmark Karatsuba multiplication followed by Montgomery Reduction on my GTX 780 Ti. My implementation of 320-bit Montgomery Reduction (CIOS) works well on the Ti, but Karatsuba + Montgomery may work better with numbers greater that 1024-bit in size.

[bsunau7]

As I think Supercomputing mentioned, using larger numbers instead of 2310*n+97, say some number 200 bits long instead of 2310, could go a long way.
Regarding the metric, I proposed "range scanned / s @ diff", but "time per 2*32 nonce @ difficulty" would probably work just as well, just don't forget to adjust for the numbers you are skipping (4 out of 5 in the 2310 case). I agree that it's difficult to compare between different difficulties...

I never liked adjusting for skipping numbers, if I can code something which only one in a million numbers are considered p6 candidates (i.e skips over 1m numbers on average) than I shouldn't have to adjust my rate by a million.  This is why the "range/s" @ difficulty seems like the best fit; clever (or not so clever) algorithms can be rated.

@SC I'd love to know how you maintain primordials of that size (you don't have to tell, but I've kept mine to less than 64bits to help with other parts of my code).

Regards,

--
bsunau7

[vinani]

I'd appreciate some feedback for one of Riecoin's possible new website.

http://riecoin.org/newsite/#home

For example:

Change background image?
Change font?
Change font color?
Change placement?
Add/Delete something?


My Riecoin address if anyone wants to give me a tip:
RRfX7TMuU9sYTJEG2RJEjEn4Z1vK3agkJM

It's nice design. From my points of view, I think it would need the navigation arrow (back and forward arrow). Because I have to scroll down to the end of the page to find the Back to Home page button. It seem to be easily to "get lost" when you are inside one page and dont scroll down to find the back button.

[steban]

I'd appreciate some feedback for one of Riecoin's possible new website.

http://riecoin.org/newsite/#home

For example:

Change background image?
Change font?
Change font color?
Change placement?
Add/Delete something?


My Riecoin address if anyone wants to give me a tip:
RRfX7TMuU9sYTJEG2RJEjEn4Z1vK3agkJM

sorry, I think it looks awful.

[remistevens]

I'd appreciate some feedback for one of Riecoin's possible new website.

http://riecoin.org/newsite/#home

For example:

Change background image?
Change font?
Change font color?
Change placement?
Add/Delete something?


My Riecoin address if anyone wants to give me a tip:
RRfX7TMuU9sYTJEG2RJEjEn4Z1vK3agkJM

I think its mostly cosmetic but, A few points:


- plain background is too plain, (we're getting part of RIC code written out on chalkboard for a background)
- don't like the body text and link fonts
- logos/video should be centered
- I think video should be on landing page - might be a personal bias
- landing page needs 1 -2 lines describing what Riecoin is. Maybe even a full paragraph. . .
- logos need different presentations in About page. Maybe section headers only use the circle logo without the word "Riecoin"? Sometimes it sits on a plate with a dropshadow? That sort of thing, needs to change up.
- all logos of Riecoin should be clickable and take you back to the landing page.
- tabs at the top might look better than big buttons. Darkcoin site for reference: top tabs, centered logo and brief description with cool background.
        http://www.darkcoin.io/
- body text backgrounds need a texture. maybe paper or wood to go with the chalkboard?
        http://pixabay.com/en/gray-notebook-paper-newspaper-page-35819/
        http://pixabay.com/en/wooden-floor-paper-texture-215877/


sorry if these make it a total overhaul

[dga]

As I think Supercomputing mentioned, using larger numbers instead of 2310*n+97, say some number 200 bits long instead of 2310, could go a long way.
Regarding the metric, I proposed "range scanned / s @ diff", but "time per 2*32 nonce @ difficulty" would probably work just as well, just don't forget to adjust for the numbers you are skipping (4 out of 5 in the 2310 case). I agree that it's difficult to compare between different difficulties...

I never liked adjusting for skipping numbers, if I can code something which only one in a million numbers are considered p6 candidates (i.e skips over 1m numbers on average) than I shouldn't have to adjust my rate by a million.  This is why the "range/s" @ difficulty seems like the best fit; clever (or not so clever) algorithms can be rated.

@SC I'd love to know how you maintain primordials of that size (you don't have to tell, but I've kept mine to less than 64bits to help with other parts of my code).

What do you mean specifically by "maintain"?  I'm happy to spill the beans on my big primorial version, since my hacked verison of jh's is faster at this point. ;-)

  -Dave

[remistevens]

Regarding the roadmap, cool! Very good as a start. I'd add Android wallet next to the Mac wallet, I've been thinking on taking on those 2. Also GPU miner.
Open source stratum-mining is my first priority now, it will get many pools and it's easier to do than p2pool, which should be later on the list. I firmly believe we have the potential to grow big once there are lots of pools online, since that will bring more decentralization and trust.
Also get everyone to accept RIC! and website and video in many languages
But don't disregard what goin2mars says: faucets, giveaways, recognition, TOR, news articles. We got an article on NewScientist! let's get more
Let's complete the article in wikipedia. Also there's a wiki here: http://www.reddit.com/r/RieCoin/wiki/index

all makes sense to me, excellent additions. thanks Gatra. I'll add these and other suggestions to make up a full list.

[steban]

Is there a paper wallet generator for Riecoin?

[remistevens]

Is there a paper wallet generator for Riecoin?

Thats a good one for the list.

I just had another thought:

- create school friendly Riecoin packages. It has client and miner, images graphs and stats, info on primes, ideas for classroom contests etc. . . I'm thinking the cool grade 4 science teacher who has the classroom lizard and hanging solar system could also run Riecoin on the homeroom computer. Teach kids about Bitcoin and Primes with Riecoin. An older student package could be offered to high schools/ universities etc for paper writing and coding projects. . . Then we just get educational email lists and start sending the packages out.

This could also be a great blow for the crypto revolution in general. Governments keep monetary policy out of schools for a reason. If people learned how the system worked at an early age, we'd never have gotten to the widespread corruption/deception we're at today. Schools would probably frown on a teacher bringing in bitcoin mining for the kids, too much like money or stock trading in most people's eyes. . .but if the teacher is bringing in software that performs an interesting mathematical task that is completely different. That would probably be encouraged. . . . sneaking Bitcoin into schools if you will. . .

Getting bitcoin into the education system will be critical eventually for its acceptance - Riecoin is positioned incredibly well towards that end.

[steban]

Is there a paper wallet generator for Riecoin?

Thats a good one for the list.

I just had another thought:

- create school friendly Riecoin packages. It has client and miner, images graphs and stats, info on primes, ideas for classroom contests etc. . . I'm thinking the cool grade 4 science teacher who has the classroom lizard and hanging solar system could also run Riecoin on the homeroom computer. Teach kids about Bitcoin and Primes with Riecoin. An older student package could be offered to high schools/ universities etc for paper writing and coding projects. . . Then we just get educational email lists and start sending the packages out.

This could also be a great blow for the crypto revolution in general. Governments keep monetary policy out of schools for a reason. If people learned how the system worked at an early age, we'd never have gotten to the widespread corruption/deception we're at today. Schools would probably frown on a teacher bringing in bitcoin mining for the kids, too much like money or stock trading in most people's eyes. . .but if the teacher is bringing in software that performs an interesting mathematical task that is completely different. That would probably be encouraged. . . . sneaking Bitcoin into schools if you will. . .

Getting bitcoin into the education system will be critical eventually for its acceptance - Riecoin is positioned incredibly well towards that end.

sounds great!

[steban]

Is difficult to understand that coins like:  

Kithecoin

Which are obvious pumps and dump, still have a higher market cap than Riecoin:

http://imgur.com/iTAWNIb

[remistevens]

Is difficult to understand that coins like:  

Kithecoin

Which are obvious pumps and dump, still have a higher market cap than Riecoin:

http://imgur.com/iTAWNIb

Kitteh lists 6 pools on their announcement page. I think that's what it comes down to. This coin isn't really decentralized until they've got it mining in several pools. Great idea or not, the mining needs to be decentralized. . . People are on it, its coming.

kittens are also cuter than Bernhard Riemann

[steban]

Rie at #59

http://crypt.la/2014/03/01/top-100-digital-currencies-by-social-media-presence/

[kingscrown]

Just posted about the coin on my blog with 1k mailing list of followers: http://fuk.io/mineable-cryptocurrencie-to-invest-in-right-now/

[surfer43]

I'm mining this interesting coin...
The fair launch will really be a motivator for more adoption...
People, don't let this crypto fall into oblivion like Datacoin  
I'm HODLING