Author

Topic: [ANN][BLC] Blakecoin Blake-256 for GPU/FPGA With Merged Mined Pools Stable Net - page 177. (Read 409569 times)

full member
Activity: 224
Merit: 100
The definition of insanity is doing the same thing
I would be also interested - maybe this will help with stratum server code....

feeleep

Its this ... pretty much works as is (just change self.ROUNDS from 14 to 8 ), though I needed to add a tiny wrapper for midstate.

I've put it at https://github.com/kramble/FPGA-Blakecoin-Miner/tree/master/MiningSoftware/blake_python for now, blake8_test.py shows how to call it. Copyright indicates free usage ...
Quote
        Copyright (c) 2009-2012 by Larry Bugbee, Kent, WA
        ALL RIGHTS RESERVED.
       
        blake.py IS EXPERIMENTAL SOFTWARE FOR EDUCATIONAL
        PURPOSES ONLY.  IT IS MADE AVAILABLE "AS-IS" WITHOUT
        WARRANTY OR GUARANTEE OF ANY KIND.  USE SIGNIFIES
        ACCEPTANCE OF ALL RISK. 

    To make your learning and experimentation less cumbersome,
    blake.py is free for any use.     
   
    Enjoy,
       
    Larry Bugbee
    March 2011

Kramble,

I recall you mentioning that you use modelsim for the fpga testing.  In my search for blake code examples, I found the below site.  Note the folder labeled "modelsim" and the multiple SHA-3 candidates at top.  There is a modelsim folder for each.

http://www.iis.ee.ethz.ch/~sha3/blake/

full member
Activity: 224
Merit: 100
The definition of insanity is doing the same thing
I stopped mining after 2 days not find a block with 2 x 7970 burning around 520 watts at .32 Kw/h
Just is too much for a coin with 0 value for me.

I'm still finding about 1 - 2 blocks per hour with 2 x 7950s and 1 x 7990.

The frequency has slowed now that the Difficulty is 8241.48538182.  Blakecoin must have attracted more miners.

sr. member
Activity: 384
Merit: 250
I would be also interested - maybe this will help with stratum server code....

feeleep

Its this ... pretty much works as is (just change self.ROUNDS from 14 to 8 ), though I needed to add a tiny wrapper for midstate.

I've put it at https://github.com/kramble/FPGA-Blakecoin-Miner/tree/master/MiningSoftware/blake_python for now, blake8_test.py shows how to call it. Copyright indicates free usage ...
Quote
        Copyright (c) 2009-2012 by Larry Bugbee, Kent, WA
        ALL RIGHTS RESERVED.
       
        blake.py IS EXPERIMENTAL SOFTWARE FOR EDUCATIONAL
        PURPOSES ONLY.  IT IS MADE AVAILABLE "AS-IS" WITHOUT
        WARRANTY OR GUARANTEE OF ANY KIND.  USE SIGNIFIES
        ACCEPTANCE OF ALL RISK. 

    To make your learning and experimentation less cumbersome,
    blake.py is free for any use.     
   
    Enjoy,
       
    Larry Bugbee
    March 2011
legendary
Activity: 1197
Merit: 1000
I stopped mining after 2 days not find a block with 2 x 7970 burning around 520 watts at .32 Kw/h
Just is too much for a coin with 0 value for me.

Something very wrong with your setup. I've been finding an average of 6 blocks per day with my 800MHash/sec Lancelot

My last two logs ...
[6 accepted, 15215 failed] 24 hours
[9 accepted, 19691 failed] 30 hours

Bluedragon:
I've found some python code for blake, got it working to calculate hash and midstate for 8 rounds if you're interested.

sure always worth looking at code examples and midstate would be most useful for the proxy as I am using the -nm switch atm

just email it to me  Cool

I would be also interested - maybe this will help with stratum server code....

feeleep
sr. member
Activity: 384
Merit: 250
Bluedragon:
I've found some python code for blake, got it working to calculate hash and midstate for 8 rounds if you're interested.

sure always worth looking at code examples and midstate would be most useful for the proxy as I am using the -nm switch atm

just email it to me  Cool

Will do, it'll also go up on my github once I've integrated it into blakeminer.py (replacing those messy system calls to midstate and checkblake). I'll also drop it into fizzisist's x6500 miner and TheSeven's Modular Python Bitcoin Miner, so we'll be able to support the x6500 and ztex boards with a proper miner. Just need to get the verilog ports working (x6500 is being a real pain, see upthread).
legendary
Activity: 1509
Merit: 1030
Solutions Architect
I stopped mining after 2 days not find a block with 2 x 7970 burning around 520 watts at .32 Kw/h
Just is too much for a coin with 0 value for me.

Something very wrong with your setup. I've been finding an average of 6 blocks per day with my 800MHash/sec Lancelot

My last two logs ...
[6 accepted, 15215 failed] 24 hours
[9 accepted, 19691 failed] 30 hours

Bluedragon:
I've found some python code for blake, got it working to calculate hash and midstate for 8 rounds if you're interested.

sure always worth looking at code examples and midstate would be most useful for the proxy as I am using the -nm switch atm

just email it to me  Cool
sr. member
Activity: 384
Merit: 250
I stopped mining after 2 days not find a block with 2 x 7970 burning around 520 watts at .32 Kw/h
Just is too much for a coin with 0 value for me.

Something very wrong with your setup. I've been finding an average of 6 blocks per day with my 800MHash/sec Lancelot

My last two logs ...
[6 accepted, 15215 failed] 24 hours
[9 accepted, 19691 failed] 30 hours

Bluedragon:
I've found some python code for blake, got it working to calculate hash and midstate for 8 rounds if you're interested.
hero member
Activity: 518
Merit: 500
hero member
Activity: 774
Merit: 500
Lazy Lurker Reads Alot
I stopped mining after 2 days not find a block with 2 x 7970 burning around 520 watts at .32 Kw/h
Just is too much for a coin with 0 value for me.
full member
Activity: 132
Merit: 100
BitCoin
stuck at block 11765, no new blocks on network for more than 30 minutes, any idea why?

11771 here so try a restart of the wallet?

if you are mining on the same machine try to reduce the intensity as it maybe taking cpu time away from the wallet

Im mining on the same machine with cgminer, i will lower the intensity , thanks!
legendary
Activity: 1509
Merit: 1030
Solutions Architect
stuck at block 11765, no new blocks on network for more than 30 minutes, any idea why?

11771 here so try a restart of the wallet?

if you are mining on the same machine try to reduce the intensity as it maybe taking cpu time away from the wallet
full member
Activity: 132
Merit: 100
BitCoin
stuck at block 11765, no new blocks on network for more than 30 minutes, any idea why?
legendary
Activity: 1509
Merit: 1030
Solutions Architect
question for developer:
if coin supply is 7 billion , 3 minute blocks , ect it seems to me Blake Coin is going to be churning out coins for a long , long time. I looked at your source and cannot tell if the extra award is calculated in this. Thought i saw max size 50 i guess if difficulty gets really large over time.
Is the difficulty subsidy counted in the 7 billion ?
Even if it is I think Blake Coin may be one of the coins with the longest ever production cycle?
(of course SIC simple inflation coin the Russian coin is designed to run forever)
Still that is a lot of coins for a long time.
I am curious why you decided on these amounts & long time span?
Not criticizing your work just curious the logic behind this?

long life cycle sounds great I do want to be able to mine Blakecoin for many many years to come don't you?

7 billion is coin Max, this includes block reward and inflation, I did not see any reason to have a low coin max and artificially create rarity it does not make sense to me  Huh

I picked 7 as it is both a symbol of luck and a prime number  Cheesy

should not be any cap on reward for Blakecoin, the idea was to create a steady coin supply that did not cut reward for miner over time and to use a fast hash function that would work on CPU/GPU/FPGA without being SHA-256 Asic compatible e.g a new main algorithm

SIC,QRK,YAK do at least try to do something different but the main thing I did not like was that they use a type of waterfall hashing from one algorithm to another which is artificially slow and would not fit in an FPGA and I was working with scrypt in FPGA with kramble's Litecoin miner but due to scrypt's linear function it was clear to me that it could never really take full advantage of the FPGA compared with the speeds of the GPU's.

I have also been working on a free to play 3D MMO framework with another developer since 2011 and thought it would be a good idea if we could use mining a coin within the game while the user was playing, "paid to pay/earn while you play" type of thing but after some research it was clear that the difficulty of Bitcoin was to high and it was not possible to merge mine scrypt based coins.

what was needed was a fast lightweight hash function that did not use too much memory and after some research the candidates where Blake-256, BMW-256, Blake2s all of which are very fast and have as much if not more security than SHA-256.

Blake-256 won for me as it is easy to work with lots of examples on CPU/GPU/FPGA and once I had done a reduced round variant was almost as fast as Blake2s and faster than BMW-256  Grin

Hence Blakecoin was born  Cool

atm still working on the pool stuff am rewriting the block submission function as it has bugs  Sad

once the pool stuff is done I will be working on the blockchain explorer and a merged coin but a kickstarter/crowd funding for the first game title is due for xmas so need some time to work on that as well, it will be the post apocalyptic FPS MMO think Mad Max, Fallout, Diablo.

I feel that Blakecoin has a really good chance at becoming the second largest cryptographic coin for CPU/GPU/FPGA in the world but it might be a little unrealistic to think that it can surpass Bitcoin in the near future but who knows maybe 10 years from now Blakecoin will be number one  Grin    

Hope that answers your questions of why Blakecoin exists and what its future is with some background on the design decisions  Cool  
  
legendary
Activity: 1148
Merit: 1018
It's about time -- All merrit accepted !!!
question for developer:
if coin supply is 7 billion , 3 minute blocks , ect it seems to me Blake Coin is going to be churning out coins for a long , long time. I looked at your source and cannot tell if the extra award is calculated in this. Thought i saw max size 50 i guess if difficulty gets really large over time.
Is the difficulty subsidy counted in the 7 billion ?
Even if it is I think Blake Coin may be one of the coins with the longest ever production cycle?
(of course SIC simple inflation coin the Russian coin is designed to run forever)
Still that is a lot of coins for a long time.
I am curious why you decided on these amounts & long time span?
Not criticizing your work just curious the logic behind this?
full member
Activity: 224
Merit: 100
The definition of insanity is doing the same thing
Robust-v03, no go.  Loads but sits idle.

Robust-v04, runs.  Loads and runs at 50MHz per core; overclockable to 63MHz; rejects shown.  Above 63MHz, reports 4294967295MHz.

Great, thanks. Well at least its some progress and we've got a working dual core now. 63Mhz should give 126Mhash/s per fpga and 252MH/s for the board, which is roughly three shares per minute or 211 per hour. Might give you around one block per day.

I can't think how to progress this further without some hands on debugging, so I think I'll call it a day there and I'll get started afresh tomorrow on my lancelot board. It'll probably take a few days to get into a position where I can make some more progress.

Thanks for all the help.
Mark


Thanks Kramble.  I'll let it run and see if it hits a block.

sr. member
Activity: 384
Merit: 250
Robust-v03, no go.  Loads but sits idle.

Robust-v04, runs.  Loads and runs at 50MHz per core; overclockable to 63MHz; rejects shown.  Above 63MHz, reports 4294967295MHz.

Great, thanks. Well at least its some progress and we've got a working dual core now. 63Mhz should give 126Mhash/s per fpga and 252MH/s for the board, which is roughly three shares per minute or 211 per hour. Might give you around one block per day.

I can't think how to progress this further without some hands on debugging, so I think I'll call it a day there and I'll get started afresh tomorrow on my lancelot board. It'll probably take a few days to get into a position where I can make some more progress.

Thanks for all the help.
Mark
full member
Activity: 224
Merit: 100
The definition of insanity is doing the same thing
Atavacron

Two more bitstreams, the 100Mhz build of the previous ...

https://www.dropbox.com/s/vrt69kzgboqf6nn/X6500-Robust-v03-2core-fmax-100MHz.bit

And one where I've added a clock buffer to the jtag clock (hopefully without breaking it) ...

https://www.dropbox.com/s/aalxmdbumare7ez/X6500-Robust-v04-2core-fmax-100MHz.bit

Both claim a fmax of 103Mhz, though perhaps I need to add some explicit constraints for hash_clk and jt_tck.

Fingers crossed, see what you make of them.

Mark

Robust-v03, no go.  Loads but sits idle.

Robust-v04, runs.  Loads and runs at 50MHz per core; overclockable to 63MHz; rejects shown.  Above 63MHz, reports 4294967295MHz.

full member
Activity: 224
Merit: 100
The definition of insanity is doing the same thing
Atavacron

Two more bitstreams, the 100Mhz build of the previous ...

https://www.dropbox.com/s/vrt69kzgboqf6nn/X6500-Robust-v03-2core-fmax-100MHz.bit

And one where I've added a clock buffer to the jtag clock (hopefully without breaking it) ...

https://www.dropbox.com/s/aalxmdbumare7ez/X6500-Robust-v04-2core-fmax-100MHz.bit

Both claim a fmax of 103Mhz, though perhaps I need to add some explicit constraints for hash_clk and jt_tck.

Fingers crossed, see what you make of them.

Mark

Got 'em.  Testin' time  Grin

sr. member
Activity: 384
Merit: 250
Atavacron

Two more bitstreams, the 100Mhz build of the previous ...

https://www.dropbox.com/s/vrt69kzgboqf6nn/X6500-Robust-v03-2core-fmax-100MHz.bit

And one where I've added a clock buffer to the jtag clock (hopefully without breaking it) ...

https://www.dropbox.com/s/aalxmdbumare7ez/X6500-Robust-v04-2core-fmax-100MHz.bit

Both claim a fmax of 103Mhz, though perhaps I need to add some explicit constraints for hash_clk and jt_tck.

Fingers crossed, see what you make of them.

Mark
sr. member
Activity: 384
Merit: 250
It looks like the max freq using "X6500-Robust-v02-fmax-100MHz.bit" is 63MHz.  If I try 64MHz the reported freq is 4294967295MHz and the miner sits idle.

OK, thanks for that. Definitely very strange. Seems setting the DCM hash_clk too fast is breaking the JTAG comms, but they should be completely independent. And 64MHz is not that fast at all, so I doubt if its a PSU issue. Some thinking is needed.

PS. Looking again at the jtag_comm code, I don't much like the fact that jt_clk is not driven from a dedicated clock buffer, and the use of the (buffered) 100MHz osc clock directly to drive dcm_progclk is also possibly suspect. I'll have a go at changing these to see if it helps.
Jump to: