Especially 16-bit ops

update build files

add depthspace, add batchspace_8

try to fix high priority issues

update copyright dates

update makefiles

stability improvements, l2pool ops

ensure we do not change a PMU counter if we never use a PMU event, allow depth padding in depthwise d32 convs

build changes for external

update copyrights

add support for 32 bit bias

code should handle different left pad values now, enable d32 LRN, mobilenet not giving correct values

add shortint dtor

Fix bugs related to 32bit bias and input conv

update with internal code base

update to internal code base

* BatchToSpace / SpaceToBatch
* DepthwiseConvolution
* Pad
* ResizeBilinear

Optimized implementations:

* Elementwise Ops (add/sub/etc)
* LRN

And various bug fixes.

Change-Id: Ia0e3da4c491c7761051b402b8c180dc9b376209a