Reports

If anyone is interested in real numbers for the fixed code:

---------------------------------------------------------------------
Benchmark                           Time             CPU   Iterations
---------------------------------------------------------------------
BM_DirectCopy/threads:1        100234 ns       102534 ns         7467
BM_AVX2/threads:1              127413 ns       125558 ns         5600
BM_MemcpyChunked/threads:1     123646 ns       122768 ns         5600
BM_Memcpy/threads:1             92502 ns        87891 ns         6400

I run it on Intel, so I replaced NEON with AVX2 equivalent.

79312464