Signed 32-bit integer elementwise addition.
Computes via a plain scalar loop because vDSP has no integer primitives. Only used by the integer-tensor add path.
Parameters
a, bconst std::int32_t*Input buffers.
outstd::int32_t*Output buffer.
nstd::size_tElement count.
Notes
Overflow wraps modulo following standard C++ semantics for signed integer addition on Apple Silicon.