[Latest Version: 3.5.0]

This is an updated version of cgminer 3.3.1 that supports mining any Scrypt, Adaptive-N-Scrypt, or Scrypt-chacha (aka "scrypt-jane") hashing algorithm.  This release includes the latest OpenCL scrypt-chacha kernel by mikaelh (v7) and a number of improvements from the version of CGMiner it was forked from (3.3.1)

Latest Version: 3.5.0 - Updated 3/13/2014

Source
https://github.com/Thirtybird/YACMiner
Windows Binaries
https://github.com/Thirtybird/YACMiner/releases

Changes:

• Display updated and fixed to account for vardiff
• ADL code has been updated and some fixes for various cards have been included
• Added fine-tuning parameters for mining high N-factor coins (Raw Intensity, Buffer-size) (starting with 3.4.0)
• Ability to mine other SC coins, and save the settings per-pool (starting with 3.4.2)
• Ability to mine Scrypt and N-Scrypt coins (starting with 3.5.0)

Usage:
sample settings for a R7 250 (384 shaders) 2GB card (with N = 14 / 32,768) being able to allocate 1,648 MB of GPU memory: 2.5 KH/sec with overclocking

Code:

yacminer --scrypt-chacha --worksize 256 -g 1 --lookup-gap 4 -R 3072 --thread-concurrency 52736 -o stratum+tcp://yac.coinmine.pl:9088 -u username -p password

Good sample settings for a 7850 (1024 shaders) 1GB card (with N = 14 / 32,768) being able to allocate 760 MB of GPU memory: 1.45 KH/sec with overclocking

Code:

yacminer --scrypt-chacha --worksize 256 --lookup-gap 8 -g 1 -R 1856 --thread-concurrency 48640 -o stratum+tcp://yac.coinmine.pl:9088 -u username -p password

At higher NFactors, cards with high shader counts should be set to use a higher lookup gap in order to keep them busy as they may be short on memory.  If you've got a 1GB card, and are mining YACoin, your lookup-gap will likely be 8.  2GB cards will probably be ranging from 4 through 8 depending on how many shaders your card has.

Parameters work similarly as in normal scrypt mining except that much more memory is needed depending on the N parameter currently used by the network. Try to find the highest value for thread concurrency you can achieve before the thread memory gets shifted from dedicated memory to synamic memory (use GPUz or HWInfo to monitor this). Set worksize (-w) to 32, 64, 128 or 256 - the RawIntensity for most cards needs to be a multiple of this value, and you can do a bit more fine tuning with a lower work size.  This may be different for the R7 & R9 cards, however, these may allow their Raw Intensity to be adjusted in multiples of 8 instead of a multiple of work size.

Your starting point for setting the Raw Intensity will be based on the amount of memory the card has.  Increase Raw Intensity until you start seeing HW errors then back it down until you aren't receiving any.  The following table is merely a guideline on where to start.

Code:

Memory	# of shaders	LG	Raw Intensity
1GB	Low Shaders	4	Shaders*4
2GB	Low Shaders	2	Shaders*2
2GB	Low Shaders	4	Shaders*4
1GB	High Shaders	8	Shaders*1
2GB	High Shaders	4	Shaders*2

Future Plans - see https://github.com/Thirtybird/YACMiner/wiki/Plan

• Ability to save configurations per card per hashing algorithm (to easily allow failover pools that are a different coin/NFActor)
• Easy-mode configuration - specify just the buffer-size and let YACMiner figure out the other settings

Attribution:
Original version by mikaelh
http://yacointalk.com/forum/index.php/topic,350.0.html
Original version by hanzac
https://bitcointalk.org/index.php?topic=232241.0