[ANN][BLC] Blakecoin Blake-256 for GPU/FPGA With Merged Mined Pools Stable Net

[BlueDragon747]

Bluedragon, what bitcoin version did you use as base?

main line 0.8.5 official release not the beta
https://github.com/bitcoin/bitcoin/tree/0.8.5

my prototype was in 0.8.3 but I only ported my changes over to 0.8.5

[kr105]

Ok thanks, I will run a diff to see if I can find any other important change.

[kramble]

I was unsure if this was due to a bug on my cgminer implementation or on the pool, now that you tested your own working cgminer and it didn't work too i will take a further look on the pool.

The only changes I made were the single sha256 in gen_hash() plus the diff1targ in test_nonce() which is actually done directly in submit_nonce in cgminer 3.1.1

The only query I have is about set_target() line 5976 (of your cgminer.c) which has ...

   if (opt_scrypt)
      d64 *= (double)65536;

I wonder if this needs to be (double)256 for blake?

I got it like this on my cgminer:
https://github.com/kR105/cgminer/blob/3.7/cgminer.c#L3140

Ah. I missed that one entirely (but it does seem to be working without it). I'll put it in and see it it makes any difference.

I'll admit that I'm rather twiddling in the dark here as I haven't analyzed the code to work out exactly what the difficulty/target code is actually doing. Though I am a little concerned that you're apparently using base 256 as the diff1 target while this was actually changed to base 0 in an early patch release of the blake wallet (if I'm recalling correctly, bluedragon may want to chip in here).

[BlueDragon747]

I was unsure if this was due to a bug on my cgminer implementation or on the pool, now that you tested your own working cgminer and it didn't work too i will take a further look on the pool.

The only changes I made were the single sha256 in gen_hash() plus the diff1targ in test_nonce() which is actually done directly in submit_nonce in cgminer 3.1.1

The only query I have is about set_target() line 5976 (of your cgminer.c) which has ...

   if (opt_scrypt)
      d64 *= (double)65536;

I wonder if this needs to be (double)256 for blake?

I got it like this on my cgminer:
https://github.com/kR105/cgminer/blob/3.7/cgminer.c#L3140

Ah. I missed that one entirely (but it does seem to be working without it). I'll put it in and see it it makes any difference.

I'll admit that I'm rather twiddling in the dark here as I haven't analyzed the code to work out exactly what the difficulty/target code is actually doing. Though I am a little concerned that you're apparently using base 256 as the diff1 target while this was actually changed to base 0 in an early patch release of the blake wallet (if I'm recalling correctly, bluedragon may want to chip in here).

I did change the shift in the wallet as it was incorrect in early release version it now matches a 29 shift like Bitcoin
https://github.com/BlueDragon747/Blakecoin/blob/master/src/rpcblockchain.cpp#L31

the shift reduced the difficulty by a multiple of 256 so you can see kramble's point 65536/256 = 256

the other value that is different to Bitcoin/Litecoin is the bnProofOfWorkLimit
https://github.com/BlueDragon747/Blakecoin/blob/master/src/main.cpp#L38

[kramble]

I did change the shift in the wallet as it was incorrect in early release version it now matches a 29 shift like Bitcoin
https://github.com/BlueDragon747/Blakecoin/blob/master/src/rpcblockchain.cpp#L31

the shift reduced the difficulty by a multiple of 256 so you can see kramble's point 65536/256 = 256

the other value that is different to Bitcoin/Litecoin is the bnProofOfWorkLimit
https://github.com/BlueDragon747/Blakecoin/blob/master/src/main.cpp#L38

The thing that's confusing me a little is that the genesis block matches to a target of 0x000000FF (it actually hashes to 0x000000BA), then when you changed the shift I used the same hash target as bitcoin 0x00000000 for the fpga mining code.

So it appears to me that we've currently got the same diff=1 target as bitcoin. But maybe I'm just confused about this.

EDIT Looking at the calc_diff() function, this has changed significantly from cgminer 3.1.1 to 3.7 and I'm going to have to have a good hard think about what's happening in there to back-port the "d64 *= (double)16777216" change.

I'm kinda tempted just to push up my current code as it seems to be working (I'm running a bit short of time today). Any thoughts or should I just leave it to tomorrow?

[BlueDragon747]

I did change the shift in the wallet as it was incorrect in early release version it now matches a 29 shift like Bitcoin
https://github.com/BlueDragon747/Blakecoin/blob/master/src/rpcblockchain.cpp#L31

the shift reduced the difficulty by a multiple of 256 so you can see kramble's point 65536/256 = 256

the other value that is different to Bitcoin/Litecoin is the bnProofOfWorkLimit
https://github.com/BlueDragon747/Blakecoin/blob/master/src/main.cpp#L38

The thing that's confusing me a little is that the genesis block matches to a target of 0x000000FF (it actually hashes to 0x000000BA), then when you changed the shift I used the same hash target as bitcoin 0x00000000 for the fpga mining code.

So it appears to me that we've currently got the same diff=1 target as bitcoin. But maybe I'm just confused about this.

same shift as Bitcoin/Litecoin from Blakecoin wallet 0.8.6

only the work limit is different
Bitcoin has 2^32
Litecoin has 2^20
Blakecoin has 2^24

[kramble]

same shift as Bitcoin/Litecoin from Blakecoin wallet 0.8.6

only the work limit is different
Bitcoin has 2^32
Litecoin has 2^20
Blakecoin has 2^24

OK, I am confused then. I'll have to go and have a look at the code and think about this.

AS I commented above (in the EDIT), do you want my to push my current 3.1.1 miner up onto github now (and dropbox for the windows binary) or wait until I've got to the bottom of what I need to do difficulty-wise?

[BlueDragon747]

same shift as Bitcoin/Litecoin from Blakecoin wallet 0.8.6

only the work limit is different
Bitcoin has 2^32
Litecoin has 2^20
Blakecoin has 2^24

OK, I am confused then. I'll have to go and have a look at the code and think about this.

AS I commented above (in the EDIT), do you want my to push my current 3.1.1 miner up onto github now (and dropbox for the windows binary) or wait until I've got to the bottom of what I need to do difficulty-wise?

happy to test the windows binary for now  

btw your FPGA code consistently gets higher shares than my GPU's maybe its just lucky code

[kramble]

happy to test the windows binary for now  

btw your FPGA code consistently gets higher shares than my GPU's maybe its just lucky code

OK its here https://www.dropbox.com/s/7nufm2dhe4csrs6/cgminer.exe

I'll update the github in the next few minutes (I need to update the README with the windows binary link as well).

PS Thanks! Lady Fate has been blowing hot and cold at me recently (like that bitcoin block would have been worth $20k if mined solo, still it'd probably have ended up as an orphan, c'est la vie!)

[kr105]

same shift as Bitcoin/Litecoin from Blakecoin wallet 0.8.6

only the work limit is different
Bitcoin has 2^32
Litecoin has 2^20
Blakecoin has 2^24

OK, I am confused then. I'll have to go and have a look at the code and think about this.

AS I commented above (in the EDIT), do you want my to push my current 3.1.1 miner up onto github now (and dropbox for the windows binary) or wait until I've got to the bottom of what I need to do difficulty-wise?

Please push your code, maybe we can use yours instead of mine as "official" cgminer for now.

[kramble]

Please push your code, maybe we can use yours instead of mine as "official" cgminer for now.

No, it does not include the OPENCL code so its only useful for FPGA mining. Your code is probably the better base for general use (and its better quality as my code is quite badly hacked around as I didn't know how to modify the makefiles at the time).

[mogrith]

Thanks
Kept the old one just in case.
Trying out the new miner on solo.

[kr105]

Keep an eye on your miners, I think I solved the issue but please confirm yourselves.

[BlueDragon747]

Keep an eye on your miners, I think I solved the issue but please confirm yourselves.

Rejected (high-hash)  

at least its a new message 

[BlueDragon747]

yay pool is working again  

Edit:
gone back to (high-hash)  

Edit2:

GPU and cm1/ztex 1.15y fpga's all working on pool nope gone back to (high-hash) again  

got a few blocks in when it was working though  

Edit3:

 

[mogrith]

Not sure yet normally find a block every 2 hrs or so.  It's over 7 hrs no block. but that's not all that out of line if it goes 18+ I'll be very worried.

More people testing means higher diff so that and normal variance can be the reason.

Anyone found a block Solo on a FPGA with the new code?

[BlueDragon747]

Not sure yet normally find a block every 2 hrs or so.  It's over 7 hrs no block. but that's not all that out of line if it goes 18+ I'll be very worried.

More people testing means higher diff so that and normal variance can be the reason.

Anyone found a block Solo on a FPGA with the new code?

just found a 4.57k solo with the new code on the ztex 1.15y will also check cm1 overnight, it also looks as good as the GPU cgminer 3.7.2 on the pool  

Edit:
and a 5.83k solo on the cm1 so the new version is working solo on both ztex/cm1  

[kr105]

I'm still looking into the possible causes of the errors, I will let you know if I get something!

[mogrith]

Found a block 9 Hrs 55 min. Now to see if I get disconnected and get orphan blocks. I hope not.

[kr105]

What next? A bounty for an exchange?