On Fri, 16 Oct 2020 at 18:18, Henri Cloetens <henri.cloetens@xxxxxxxxxx> wrote: > > Hello Jonathan, > > No, it is not. It seems a gcc problem. Well you asked about building newlib without double support, so that's what I replied to. > It has to do, that even on a machine without floating point, the compiler > is supposed to support certain floating-point operations. > I think it is the calling convention - meaning the calling convention > needs to > be defined for floating point, and the processor needs to be able to > handle this. > > Next, if the floating point is absent, the compiler will use built-ins > for the floating point operations. > Q. Is there anybody having experience with this ?. Its just my guess, it > would be nice to > have some extra opinions. Have you looked at what other targets do when GCC is configured using --with-float=soft ?
