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

[bsunau7]

so you're really mining with ARM? cool!
well, thermal throttling means it's not literally cool, but you get the point

Aside from the first 2 weeks when I had a dozen or so AWS instances everything I've done has been ARM based.  Not sure I'd call it mining yet, but it is fun.

No hardware division, slower clock, really slow memory access and 32bit architecture are the downsides.  On the upside, very low power and a super scaler architecture and my intel box isn't making noise.

Anyway, I rewrote the code to make use of neon instructions; parts of the chip were are normally in a powered down state are powered up drawing power and making heat.  I replaced the stock fan with a 12v silent one and knocked 23c off the temperature (107c to 84c).

If I can get the speed to something usable was thinking about a binary release, I am amazed at how many people have little raspberry pi's doing nothing (or very little).  Might make the few 100 RIC I own a little more valuable (or might not).

aamarket with android port dga'a miner probably was thinking the same general game plan.

Regards,

--
bsunau7

[1713892300]

1713892300

[1713892300]

1713892300

[1713892300]

1713892300

[gatra]

so you're really mining with ARM? cool!
well, thermal throttling means it's not literally cool, but you get the point

Aside from the first 2 weeks when I had a dozen or so AWS instances everything I've done has been ARM based.  Not sure I'd call it mining yet, but it is fun.

No hardware division, slower clock, really slow memory access and 32bit architecture are the downsides.  On the upside, very low power and a super scaler architecture and my intel box isn't making noise.

Anyway, I rewrote the code to make use of neon instructions; parts of the chip were are normally in a powered down state are powered up drawing power and making heat.  I replaced the stock fan with a 12v silent one and knocked 23c off the temperature (107c to 84c).

If I can get the speed to something usable was thinking about a binary release, I am amazed at how many people have little raspberry pi's doing nothing (or very little).  Might make the few 100 RIC I own a little more valuable (or might not).

aamarket with android port dga'a miner probably was thinking the same general game plan.

Regards,

--
bsunau7

very interesting!
I used to work with an ARM chip that had an "embedded cryptographic coprocessor", which basically means it had a few interesting features implemented by hardware, like sha2 and modular exponentiation (up to 2048bits in this case). Wish I could put my hands on one of those, but the development boards were expensive.

[bsunau7]

very interesting!
I used to work with an ARM chip that had an "embedded cryptographic coprocessor", which basically means it had a few interesting features implemented by hardware, like sha2 and modular exponentiation (up to 2048bits in this case). Wish I could put my hands on one of those, but the development boards were expensive.

Not wanting to distract you from riecoin but look at http://www.zedboard.org/product/microzed dual core with FPGA (parallela also sell one but they've had big issues with their kickstarter campaign).  Xilinx even give you access to a (cut down) version of their development SDK.  With some free VHDL tools you should be able to generate a bit stream to define your own co-processor.  Parallela even use this to interface into their 16/64 core coprocessor and to define their HDMI hardware (yes, HDMI is defined in SW just add some hardware to support the physical interface).

Also there are some OpenCL to VHDL converters, GPU coins with simple algorithms and some market liquidity are about to get very cheap/easy to move to FPGA based mining rigs.  This is why I am mostly ignoring AES/SHA/CRYPT/X11 based coins...  A few $$$$ and some time could all but destroy them.

Regards,

--
bsunau7

[mumus]

very interesting!
I used to work with an ARM chip that had an "embedded cryptographic coprocessor", which basically means it had a few interesting features implemented by hardware, like sha2 and modular exponentiation (up to 2048bits in this case). Wish I could put my hands on one of those, but the development boards were expensive.

Not wanting to distract you from riecoin but look at http://www.zedboard.org/product/microzed dual core with FPGA (parallela also sell one but they've had big issues with their kickstarter campaign).  Xilinx even give you access to a (cut down) version of their development SDK.  With some free VHDL tools you should be able to generate a bit stream to define your own co-processor.  Parallela even use this to interface into their 16/64 core coprocessor and to define their HDMI hardware (yes, HDMI is defined in SW just add some hardware to support the physical interface).

Also there are some OpenCL to VHDL converters, GPU coins with simple algorithms and some market liquidity are about to get very cheap/easy to move to FPGA based mining rigs.  This is why I am mostly ignoring AES/SHA/CRYPT/X11 based coins...  A few $$$$ and some time could all but destroy them.

Regards,

--
bsunau7

I just got my parallella 2 days ago. I don't even remember when I ordered. Anyway it's time to play with it and find some primes . If anybody has some links to start with, I will thank them. I'm a noob in this territory.

[bsunau7]

I just got my parallella 2 days ago. I don't even remember when I ordered. Anyway it's time to play with it and find some primes . If anybody has some links to start with, I will thank them. I'm a noob in this territory.

You lucky guy (or girl)!  The 16 or 64 core epiphany one?  The cores should be perfect for prime number grinding which is why I was looking at them in the first place... 

I'll assume you are comfortable with development (otherwise why get a parallela?) so here is a quick high level road map...

The riecoin wallet compiles pretty cleanly, I'd probably start with that as it gets most of your build dependencies in place (also gets you used to linaro/ubuntu package system if it is new to you).  I used the latest dbm libraries (not the 4.8 ones recommended) as I don't need a transportable wallet.

Following that the cpu miner should compile (once again it has its own dependencies).  Performance will suck, but you'll get something up and running.  dga's miner should also work (I've not compiled it recently however).

Then tune/code/tune/code until you get a block.  Ask if you need any more help.

Regards,

--
bsunau7

[mumus]

I just got my parallella 2 days ago. I don't even remember when I ordered. Anyway it's time to play with it and find some primes . If anybody has some links to start with, I will thank them. I'm a noob in this territory.

You lucky guy (or girl)!  The 16 or 64 core epiphany one?  The cores should be perfect for prime number grinding which is why I was looking at them in the first place... 

I'll assume you are comfortable with development (otherwise why get a parallela?) so here is a quick high level road map...

The riecoin wallet compiles pretty cleanly, I'd probably start with that as it gets most of your build dependencies in place (also gets you used to linaro/ubuntu package system if it is new to you).  I used the latest dbm libraries (not the 4.8 ones recommended) as I don't need a transportable wallet.

Following that the cpu miner should compile (once again it has its own dependencies).  Performance will suck, but you'll get something up and running.  dga's miner should also work (I've not compiled it recently however).

Then tune/code/tune/code until you get a block.  Ask if you need any more help.

Regards,

--
bsunau7

Thanks for the quick answer! I have the 16 core one. I used to develop under windows (.net) but I'm starting to feel home also under linux (x86) environment. I played a lot with the code of the primecoin miner when it started, and that's why I ordered the parallella back then. ( I've checked, in October 2013). I'll post some updates if I'll get it up and running with Riecoin.

[wakawakalj]

Has there been any progress on the windows version of the b15 DGA miner?

[gatra]

Has there been any progress on the windows version of the b15 DGA miner?

I compiled it with mingw64, using gmp6, and got this (in windows):

[00:03:28] 2ch/s: 0.0000 3ch/s: 0.0000 4ch/s: 0.0000 Shares total: 0 / 0
[00:03:36] 2ch/s: 0.0000 3ch/s: 0.0000 4ch/s: 0.0000 Shares total: 0 / 0
[00:03:44] 2ch/s: 0.0000 3ch/s: 0.0000 4ch/s: 0.0000 Shares total: 0 / 0
[00:03:52] 2ch/s: 0.0000 3ch/s: 0.0000 4ch/s: 0.0000 Shares total: 0 / 0
[00:04:00] 2ch/s: 0.0000 3ch/s: 0.0000 4ch/s: 0.0000 Shares total: 0 / 0
[00:04:08] 2ch/s: 0.0000 3ch/s: 0.0000 4ch/s: 0.0000 Shares total: 0 / 0
[00:04:16] 2ch/s: 0.0000 3ch/s: 0.0000 4ch/s: 0.0000 Shares total: 0 / 0
[00:04:24] 2ch/s: 0.0000 3ch/s: 0.0000 4ch/s: 0.0000 Shares total: 0 / 0

same source code worked on linux. Will try to compile in visual studio in order to debug with its IDE

[aamarket]

Hi all Riecoiners !

@bsunau7
Android port of DGA's miner was just a side note, I prefer Linux to Android.

I checked some net info about FPGA modular exponentation performance, nothing interesting or too expensive,
but possibly somebody has better info.

@mumus
Paralella can hardly be any better, I think they dropped the version with more interesting fpga chip, but I am interested in real 16 core results
DGA's miner should compile and work directly, he corrected the problem i pointed out. (search history here).
Should you need a help, let me know.

but too little time left, I just comment the diff occasionally,

1395 now is looooooooong time low, I contemplate about starting the unoptimized ARM miner again

[gatra]

1395 now is looooooooong time low, I contemplate about starting the unoptimized ARM miner again

ypool had some problems last night, which caused diff to drop a little
also, I think some cpu miners moved to new cpu coins with cryptonight and maybe "wild keccak" (boolberry), but gpu miners exist for both so I expect them to come back soon

[mumus]

Hi all Riecoiners !

@bsunau7
Android port of DGA's miner was just a side note, I prefer Linux to Android.

I checked some net info about FPGA modular exponentiation performance, nothing interesting or too expensive,
but possibly somebody has better info.

@mumus
Paralella can hardly be any better, I think they dropped the version with more interesting fpga chip, but I am interested in real 16 core results
DGA's miner should compile and work directly, he corrected the problem i pointed out. (search history here).
Should you need a help, let me know.

but too little time left, I just comment the diff occasionally,

1395 now is looooooooong time low, I contemplate about starting the unoptimized ARM miner again

I've succeed to compile DGA's miner on the Parallella without any problem and the speed on the main dualcore ARM A9 CPU is 0.3655 2ch/s. Not impressive but I didn't expected more. The problem is with the share submission. In the moment it finds a share and trying to submit it, xptMiner crashes and I get the error message "Bus error". I didn't had the time to debug it yet. I think it's some alignment problem.
When I'll have some free time I will also start to port the code to use the 16-core Epiphany chip, just for fun. I don't expect to see any dramatic speed improvement.

[bsunau7]

I've succeed to compile DGA's miner on the Parallella without any problem and the speed on the main dualcore ARM A9 CPU is 0.3655 2ch/s. Not impressive but I didn't expected more. The problem is with the share submission. In the moment it finds a share and trying to submit it, xptMiner crashes and I get the error message "Bus error". I didn't had the time to debug it yet. I think it's some alignment problem.
When I'll have some free time I will also start to port the code to use the 16-core Epiphany chip, just for fun. I don't expect to see any dramatic speed improvement.

Your right the problem is the xpt data packets are not aligned, everything (except neon) is 32bit aligned in ARM.  To debug it I had to compile without -03 (makes the code flow easier by also slows you right down) and gdb it.  Some ugly alignment code (using arm load/stores or array assignments) should work, once you know what needs to be fixed.  jh00 knows about this.

On performance, compile and use gmp 6.0 if you aren't already using it.  I saw a ~14% (from memory) improvement in my code moving to that version.

Regards,

--
bsunau7

[bsunau7]

Hi all Riecoiners !

@bsunau7
Android port of DGA's miner was just a side note, I prefer Linux to Android.

I checked some net info about FPGA modular exponentation performance, nothing interesting or too expensive,
but possibly somebody has better info.

I would tend to agree with you on the FPGA; the (highly) variable difficulty (integer length) or riecoin would make anything which operated on the whole integer very difficult/expensive on an FPGA.  Another approach would be a (fixed integer length) sieve in the FPGA and keep the difficult bits on a CPU.

Problem with that (and the one I am having) is you wind up hand coding an optimized sieve which has a scale out issues both in terms effort/reward and gates on the chip.  Each tested prime seems in the sieve knocks off ~10% of the candidates, so quickly sieving the first few dozen primes should should see the bulk of the gains.

As I said this is more of an idle fun project than earning coin.

Regards,

--
bsunau7

[SpeedDemon13]

So, there still isn't a gpu miner for RIC? Or at least a more improved cpu miner?

[aamarket]

I've succeed to compile DGA's miner on the Parallella without any problem and the speed on the main dualcore ARM A9 CPU is 0.3655 2ch/s. Not impressive but I didn't expected more. The problem is with the share submission. In the moment it finds a share and trying to submit it, xptMiner crashes and I get the error message "Bus error". I didn't had the time to debug it yet. I think it's some alignment problem.
When I'll have some free time I will also start to port the code to use the 16-core Epiphany chip, just for fun. I don't expect to see any dramatic speed improvement.

Hi, I had the error as well, and the speed was similar (*2 for 4core) - see
https://bitcointalk.org/index.php?topic=446703.msg6718714#msg6718714
I am surprised by what you say about 16-core chip ; what's the reason to buy that device at all then ?


@SpeedDemon13 : afaik no public one

[GambitBTC]

Take RIC for what it is, a very valuable experiment.


Crypto as a whole is moving forward.

[northranger79510]

Take RIC for what it is, a very valuable experiment.


Crypto as a whole is moving forward.

RIC is still valuable. Any crypto with improvements and active dev. is a good crypto.

[northranger79510]

BTW Gatra, any news after you release 0.9.2? update Thursday here once again!

[GambitBTC]

Take RIC for what it is, a very valuable experiment.


Crypto as a whole is moving forward.

RIC is still valuable. Any crypto with improvements and active dev. is a good crypto.

I agree, but this truly has no chance of being adopted by the public (like 99% of alts) but nonetheless has provided a few unique concepts that can help us move along as a community.

[cinnamon_carter]

dont be so sure riecoin cannot be valuable in the long run and used by mainstream people

i have a question, from research i notice ypool hits almost every block

we have all heard the drama over gigahashio and bitcoin and the stories of block witholding attacks against other pools

can someone tell me how that is not an issue here ?

not only is it one pool it is closed source,

i would not run anything closed source ever, anyone who does should reconsider