* check compiler support
  * check HW support before executing
  * add test doing round trip conversion from / to FP32
  * treat array correctly if size is not multiple of 4
  * add declaration to prevent warning
  * make it possible to enable fp16 on 32bit ARM
  * let the conversion possible on non-supported HW, too.
  * add test using both HW and SW implementation

- added new functions from core module: split, merge, add, sub, mul, div, ...
- added function replacement mechanism
- added example of HAL replacement library

IPP_VERSION_MAJOR * 100 + IPP_VERSION_MINOR*10 + IPP_VERSION_UPDATE
to manage changes between updates more easily.

IPP_DISABLE_BLOCK was added to ease tracking of disabled IPP functions;

fixed trailing whitespaces
renamed files *vaapi* to *va_intel*
changed names vaapi* to va_intel*

fixed whitespaces
fixed issues
fixed issue
fixed module statement issues
fixed access mode
added initialization check
fixed warning

Some code style corrections;

Removed IPP port for tiny arithm.cpp functions

Additional warnings fix on various platforms.

Build without OPENCL and GCC warnings fixed

Fixed warnings, trailing spaces and removed unused secure_cpy.

IPP code refactored.

IPP code path  implemented as separate static functions to simplify future work with IPP code and make it more readable.

removing trailing whitespaces

Compilation error on Mac fix & warning on android

Warnings fixed on iOs

Fixed getContinuousSize() to handle huge matrices properly. This should solve http://code.opencv.org/issues/3232

fixed http://code.opencv.org/issues/3935
possibly fixed http://code.opencv.org/issues/3594

Moved out of TLS to properly collect data from different threads;
Concurrent access guards were added to collection functions;

IPP can be switched on and off on runtime;

Optional implementation collector was added (switched off by default in CMake). Gathers data of implementation used in functions and report this info through performance TS;

TS modifications for implementations control;

(cherry picked from commit 4f9c081d)

See 092f916d for explanation.

This reverts commit 4f9c081d.

That commit introduces problems, as it breaks certain use cases of
OpenCV functions; for example, convertTo of an Nx1 matrix to an
std::vector.

Since vectors can't store separate width and height values, OpenCV
considers them to always be 1xN. So even though the vector is created with
.create(N, 1), the Mat passed to getContinousSize has dimensions 1xN,
and the size comparison fails, even though the operation itself is safe.

This is a use case we probably don't want to break, at the very least
for backwards compatibility. So I'm reverting the commit. This will also
unfix bug #3319; I'll submit a less intrusive solution as a PR to 2.4, which
will also revert 4f9c081d there.