In class MmaTensorOpWmmaMultiplicandTileIterator, the function operator++() is compute byte_offset with: byte_offset_ += (elements_offset * sizeof_bits::value) / 8; inL485
but in the function operator--(), compute byte_offset with: byte_offset_ -= (elements_offset + sizeof_bits::value) / 8; inL496
the operator++() compute with multiply, but operator() compute with add, Why?
#694 (comment)
In class MmaTensorOpWmmaMultiplicandTileIterator, the function operator++() is compute byte_offset with: byte_offset_ += (elements_offset * sizeof_bits::value) / 8; inL485
but in the function operator--(), compute byte_offset with: byte_offset_ -= (elements_offset + sizeof_bits::value) / 8; inL496
the operator++() compute with multiply, but operator() compute with add, Why?
#694 (comment)