Ok so... it has been a suffered childbirth but at last I can announce...
version 801 of my someday-will-be-complete M8M minerfor those who read about it for the first time, M8M was born as an attempt in learning OpenCL and modern GPU programming (it really started as an exercise in DirectCompute, but it sucked). It eventually become useful, back when people mined LTC with GPUs. I later discovered Myriad and decided it was a great playground so I could test multiple algorithms with a single wallet. Because I'm lazy sort of. And short on disk.
I decided it would become a miner for newcomers. It doesn't come with a manual because it hopefully doesn't need one.
It also tries to be less invasive to your work for all those users who currently do something with their PCs.
In theory, it should "just work" but in practice I only have a Radeon 7750 so the testing I can do is very limited.
This release comes with high hopes!
- As a start, M8M is now almost serious: it now contains extensive validation to detect so called "hardware errors" and filter them out;
- a new, simplified, data driven work dispatch architecture has been deployed in an attempt to move forward the whole system and hopefully improve compatibility with multi-gpu systems (which so far were always hit-or-miss);
- fixed a bug which caused difficulty changes to produce a stream of rejects;
- a first attempt at computing device efficiency akin to "work utility" on per-device and per-pool basis (but currently only shown at device level);
- nearly all issues have received some work, in particular it should be now possible to use even devices from multiple vendors (even though it doesn't make sense as algos are tested only on AMD GCN and require it) as well as not being limited by slowest card;
- ?
The kernels are the same as the previous version (+ validity checking) and should produce more or less the same hashrate.
Testers wanted! 
