On 12/01/2015 04:28 PM, Alexander Monakov wrote:
Bernd, is your position on exposing shared memory as first-class address space on NVPTX subject to change? Do you remember what middle-end issues you've encountered when trying that?
TYPE_ADDR_SPACE does not reliably contain the address space. Patches to deal with that (rather than fix it which Joseph doesn't like) got really ugly and I gave up on it. So please use the patch I sent which deals with .shared inside the ptx backend (although I think it may have to be reworked a little since Nathan changed the code around recently).
Bernd