remove the gpu_thread_num safety

Keep consistency between class and struct

fix cuda 7.5 compile

fix CMake find package output

fix error: `identifier "__builtin_ia32_monitorx" is undefined`

remove early returns

improve AMD auto suggestion

Minor doc fixes

fix #3

- remove QUIET from find package for OpenCL and CUDA
- remove pascal architecture from default CUDA_ARCH for CUDA < 8.0

- use always a multiple of the number of compute units
- add more device information to the config file

Early returns within a kernel can have unexpected behavior, this strongly depends on the opencl
runtime compiler. To avoid errors all early returns are removed.
There is no negative effect(e.g.performance) if the threads stay alive up to the end of the kernel.

One source for the early return is: http://al-key-opencl.blogspot.de/2014/09/be-careful-not-to-mix-early-return-and.html

This pull request also fix a race condition where all fill the shared memory.
Also a bug introduced with #16 is fixed, because of the early return not was possible that
the last block works with an wrong initilized shared memory (result should be a wrong hash if the result target size is valid).

avoid possible illegal memory access

fix travis OSX build

add `brew update` to avoid that `brew tap homebrew/science` crash

fix wrong timer for stats

fix #21, hash values randomly not available.
The reason is that I changed the lables but forget to change all timer.
A GPU needs sometimes ore than 2.5s to finish one hash round, therefore
the results are only visible randomly.

Is is possible that the number of results are greater than the result output array,
in this case invalid memory can be access within the device and on the host side.

fix invalid kernel call parameter

Port the crash fix from xmr-stak-cpu

check that the number of global threads is a multiple of the workgroup size

Fix large pages on Windows and make them more user-friendly

Increase the scratchpad size to make sure we don't malform JSON

fix amd auto suggestion

fix broken nicehash support

Major cleanup and speedup of soft aes

The number of threads within a kernel must be a multiple of the worksize.
If not it can crash on some systems.

Platform index zero was always used to search for the device detection.

Fix: use correct detected AMD platform index

Merge branch 'topic-large-pages-win' of https://github.com/fireice-uk/xmr-stak into topic-large-pages-win

Topic cleanup