Topic: [ANN][BLC] Blakecoin Blake-256 for GPU/FPGA With Merged Mined Pools Stable Net - page 171. (Read 409641 times)

There was a change in the way difficulty is reported in one of the early patch releases. The original client/wallet used a difficulty base of 2^24 (compare bitcoin at 2^32 and litecoin at 2^16). This was changed to 2^32 by the patch release (just cosmetic really, but under the old system we'd be approaching a diff of 1 Million which just looked nasty IMHO). Possibly the cgminer version you have is reporting on the old difficulty scale. As long as your blocks are being accepted its not really a problem.

Maybe a stupid question, but why do those numbers differ?

cgminer reports: Network diff set to 324K
getinfo (performed against the wallet): 'difficulty' => '1264.37669219',     1:256

In the past I did observe

cgminer: Network diff set to 546K
getinfo: 'difficulty' => '2133.63558956',     1:256

So there seems to be a constant of 256 involved.  network diff / getinfo diff = 256

Idea for a client improvement:

Output the 'newmint' => 'xx' value with the "getinfo" RPC call.
I saw that in other implementations and I did really like it!

I activated my old GPU setup. A Win 8 installation with HD7870 + HD7850.
Using cgminer I get 1,5GH/s with just the HD7870 and 2,5GH/s with both. No tuning done (besides
memory clockrate to min and a little bit overclocking).

No orphan blocks seen (but can't say for 100% that had been none).

On a different subject,

Kramble I got my CM1 today. So I'm getting drivers and stuff so I blindly fail around with stuff I have no idea how to use. (oddly my idea of fun  )

I'm getting a lot of orphan blocks.

mined 7 blocks so far today 3 of which or no good. using 0.8.8.8 wallet and the Cgminer 3.3.1 modded for blake.
anyone else?

Not right now, but I could build one. The ztex website says the Ztex-1.15x is compatible with the Ztex-1.15d bitstream, and there is a version of the bitcoin verilog source available for this so it shouldn't be too difficult (famous last words, I'm not at all convinced the source code actually matches the officially released bitcoin bitstreams, but the 1.15y source did work, after a fashion). I can't test it myself so it will be a blind port, but that hasn't stopped me before

Give me a few days as I've just started on the Cainsmore CM1 port and I want to make some progress on that one first.

I have it running (ok, not running) on my r-pi. But I only have two Ztex-1.15x. So the bitstream is not working.
Do you have a bitstream for the single (non quad) Ztex-1.15x?

I also have one Ztex-1.15b but this one never did work on the r-pi (is not detected by "lsusb").

Thanks, I've updated https://www.dropbox.com/s/xa01f9hhakpsexv/cgminer-3.1.1-blakefpga.zip to include it, and tweaked runblake.bat slightly. Oddly it does not seem to be needed on either of my Win7-64 or Win8-32 boxes, but I do have 7zip installed which may explain it (it will have already installed the library).

I haven't actually built this exact version on raspi, but the previous driver-icarus version did build OK, though its a standard dynamic build, my expertise does not stretch to modifying the makefile for static linking. Its been a while since I originally set up the dependencies for this, so I'm not quite sure exactly what you will need, the build instructions in the original cgminer README are a bit lacking. I'll have a dig around and see if I can find any notes I may have made ... nope, lost in the mists of time, but google did find something http://learn.adafruit.com/piminer-raspberry-pi-bitcoin-miner/install-cgminer seems straightforward enough.

When I try this on a Win8 system, I get an error about zlib1.dll missing.
Did you forget to include this library in your zip file?

Do you have a static compiled version for a Raspberry Pi?

trepex

nice
New job's for my fpga farm.

An experimental port of the FPGA Blakecoin miner to the Ztex 1.15y is now available.

Bitstream at [initial version redacted - faster versions now include pll_stop functionality]
Faster version https://www.dropbox.com/s/qlsbvarijcvlfpo/ztex_ufm1_15y1-v06ad-2core-ucf-120MHz-fmax-125.bit
Faster version https://www.dropbox.com/s/jafj67j6gmll97x/ztex_ufm1_15y1-v06ad-2core-ucf-140MHz-fmax-147.bit
Bugfix version https://www.dropbox.com/s/vk3k5sb64b8641o/ztex_ufm1_15y1-v06ad-2core-ucf-140MHz-fmax-147-fixed.bit

Cgminer patches at https://github.com/kramble/FPGA-Blakecoin-Miner/tree/master/cgminer/cgminer-3.1.1

See the README for compilation instructions and a pre-compiled windows build (feedback would be appreciated as to whether I've included all the dependencies for this). Note that you will need to disable GPU mining via --disable-gpu as it is not compiled in, and that the --debug switch may cause it to crash (it does for me). You can (and should) redirect the log output via 2>log.txt for later inspection. Some more detailed instructions here.

WARNING This is experimental. Use it at your own risk. IT MAY DAMAGE YOUR ZTEX BOARD. At the very least, monitor for overheating as the overclock and overheat protections of the original bitcoin code are not currently operational.

On the two boards its been tested on so far, it will only load up on three of the four FPGA devices. I suspect this may be an issue with the default firmware, but it seems fairly harmless. Max speed on three devices is reported at around 1GHash/sec. Use the bugfix version which works on all four devices.

Automatic clock speed management is enabled and the clock will ramp up until hardware errors occur, then back off a little. At least that's what its supposed to do. In practice it ramps up until the FPGA locks-up, reloads the bitstream and backs off one step. You will see several "checknonce failed" messages as the clock reaches maximum. Don't fret as they should settle down pretty soon after. Its a work in progress, any help from the pro's to get this working properly would be appreciated.

Have fun

How do I get the p2pool to calculate the checksum correctly? Currently I use:


The error:


The checksum is a single BLAKE(256) iteration, right? Do I have to do anything to the data (e.g. reverse it) before I send it out?

That's another reason why it hasn't hit a market. Most markets at least want a block explorer, but most want both block explorer and pool. Plus no services online yet to use the coin.

there must be endianess problem.
you can try to play with preparation step in cgminer ( findnonce.c, precalc_hash_blake256() ), there calculated midstate, and it not portable for BE/LE platforms.
also when cgminer prepare argument for kernel there also may be endianess problem.

need some pools

It seems that no one buys Blakecoin, I tried to sell some BLC at different prices, and no one was not interested in buying. I think it's because of pool\block explorer absence

No, that's definitely not hashing properly. Hash7 should be 00000000 for valid hashes. Unfortunately I know almost nothing about the GPU code in cgminer so I'm not going to be of much help. One thing to check is if this is just an occasional error or if all of the hashes coming from the GPU are invalid (the message comes from submit_nonce() in cgminer.c so just log the valid ones as well as the invalids).

any pools yet

kramble,

thanks for the tip, after messing with the source code files for about 2-3 hours I finally got cgminer to compile on 64bit linux.  i haven't touched C/C++ in like 5 years, so it's a hack job but it runs however, I am getting continuous errors like this
or like this
depending on what settings I use. is this because its not hashing properly, or is this how cgminer tells you you did not find a block? I copied the blake256.cl file from the windows cgminer archive, but i'm not sure its loading properly.
any more ideas would be helpful. Once I can confirm this build is running properly, I will upload the binary and source.