commit 152c2a36a10a14dc9c931ec62654d6bc3158863a
author Michael Butler <butlermichael@google.com> Mon Apr 26 13:47:15 2021 -0700
committer Michael Butler <butlermichael@google.com> Fri May 14 16:25:56 2021 -0700
tree 0afcda7c706ee60dcfec9e19895f0c97b38e5ed0
parent 70fdbc46231a3cd6ea9d4d3ba491534a8bef02d7

Change OperandValues' AlignedData to uint8_t

Prior to this CL, Model::OperandValues ensured the data was aligned by
using a vector of aligned data behind the scenes. Specifically, the
AlignedData was std::max_align_t, ensuring it was at least as aligned as
any primitive type used in the NNAPI. This CL changes the underlying
data to use a vector of uint8_t which is simpler to manage, but ensures
the data is still aligned by verifying __STDCPP_DEFAULT_NEW_ALIGNMENT__
is sufficient.

Bug: N/A
Test: mma
Test: NeuralNetworksTest_static
Change-Id: Idd5951f0caa7cae9b2ee990caa8cfd03c9176248