[ANN][BURST] Burst | Efficient HDD Mining | New 1.2.3 Fork block 92000

Very, very nice work

I'll try it on my linux box, but any chance for win versions...? Would be awesome Grin

Quote from: dcct on September 04, 2014, 07:52:35 PM

I like to share an updated version of the Linux C tools. You can download it here:

https://bchain.info/dcct_miner.tgz

First you need to unpack it:
tar -xvzf dcct_miner.tgz

Then compile it:
make

of if you use 32-bit Linux:
make linux32

If you really can't compile it use the ones in "bin".

Now you have 3 tools:

The plot merge utility

If you created your plots with a small stagger size, processing them takes a long time. You might miss your deadline if its found too late!
The merge utility optimizes your already created files making them a lot faster to read.

How to use it?

Start it with one parameter, the input file.

Code:

./merge plots/1234567890_0_100000_1000

This will read your plotfile and create another one - 1234567890_0_100000_100000 - in your current directory. It now has a stagger size of 100000 which is a LOT better.
When done, delete the input file and put the optimized plot in its directory.
You can now use the optimized file with both the JAVA and the C miner.

This does not work for incomplete plotfiles.

The C plotfile generator

This tool creates plotfiles. It takes the same parameters as the JAVA plotfile-generator.

Start it like this:

Code:

./plot <public key> <start nonce> <nonces> <stagger size> <threads>

The 32-bit version can handle a stagger size of about 15000, lots more (depending on your memory) with the 64-bit version.

Some recommended stagger sizes:
2GB memory -> 4000
4GB memory -> 8000
8GB memory -> 20000
16GB memory -> 50000
32GB memory -> 100000

It creates a directory "plots" and puts your files there. They exactly match the files generated with the JAVA-Tool and can be used with any miner.

The miner

You want to use your plotfiles to generate coins, this is where the miner is used.

The C version has some advantages:
- It uses almost no memory
- A lot faster, multithreaded
- Can handle multiple plot directories

You first have to create a file called "passphrases.txt" and put your passphrase in it. It supports only one passphrase and uses the files first line.
Avoid spaces before/after your passphrase.

Then start the miner:

Code:

./mine <node ip> [<plot dir> <plot dir> ..]

For example:

Code:

./mine 127.0.0.1 /mnt/disk1/plots /home/user/plots /mnt/usb1/plots

Each directory is read by a separate thread. If you use multiple HDD's try to use one directory per HDD, this is fastest.

The miner creates some nice output:

Code:

2872 MB read/11488 GB total/deadline 50134s (49664s left)
New block 8674, basetarget 9388869
2872 MB read/11488 GB total/deadline 21573s (21476s left)
New block 8675, basetarget 9260523
2872 MB read/11488 GB total/deadline 3040s (2823s left)
New block 8676, basetarget 9461804
2872 MB read/11488 GB total/deadline 1122s (1018s left)
New block 8677, basetarget 9502439

When seconds left approach zero, you likely found a block.

If you like it - you can do me a favor and use/spread my blockchain-explorer http://bchain.info. It does not support BURST unfortunately.

2872 MB read/11488 GB total/deadline 50134s (49664s left)
New block 8674, basetarget 9388869
2872 MB read/11488 GB total/deadline 21573s (21476s left)
New block 8675, basetarget 9260523
2872 MB read/11488 GB total/deadline 3040s (2823s left)
New block 8676, basetarget 9461804
2872 MB read/11488 GB total/deadline 1122s (1018s left)
New block 8677, basetarget 9502439