From: Gavin Shan > Sent: 08 May 2017 01:20 ... > >Why 16 bits? > >You are just making life hard for the compiler and possibly generating > >random padding. > > > > It's because there are 256 NCSI channels to maximal degree. So 16-bits > is the minial data width to hold it in signed format. Yes, I think > __s32 would be better in this case. However, I would like to discard > the negotiation mechanism in next respin.
Just because the domain of a value fits in 16 bits doesn't mean that a 16bit type is appropriate. It is generally much better to use 32 (aka machine word) sized items unless you have an array or are trying to fit a lot of items into a small memory area. David