Fix build warnings

Fix a bug where results were wrong if max work group size was restricted to <= 64

Fix a bug in the Twist kernel where global and local work dimension calculations were swapped.

Fixed a warning in fft_execute.cpp where the ‘inPlaceDone’ flag may be used before it is initialized.

Fix a failures due to the requested local dimension exceeding the value returned by clGetKernelWorkGroupInfo on some GPUs.

This sample demonstrates how to use OpenCL and the GPU to compute Fast Fourier Transforms.