changes are changes

[sunnyhellious]

Hi!

I have added some basic optimisations for 5% additional goal.

Actually now kernel calculates several hashes per one flight. This parameter at the moment is defined in sha256.cpp. HPF define is currently setted to 3. Hpf - hashes per flight.

And can you fix this lines for old GPU compatibility in OpenCL kernel. Atomic add operation breaks kernel work on hd4890 for instance:

// if (atomic_add(&result->found, 1) == 0)
result->found++;

[DelphiR]

The HFP is the main and fundamental feature of this PR optimisation
It can be between 3 and 100
The most omptimum quantity of it depends on the power of GPU card
For example for GTX1080Ti the HPF might be the 100 and increase the total power of algo up to 10% for example. Which is very cool!

[DelphiR]

During the AMD test session we see more effect in terms of more accel from this commit