On Sat, 2007-10-27 at 12:54 +0200, Richard Guenther wrote: > On 10/27/07, Robert Dewar <[EMAIL PROTECTED]> wrote: > > skaller wrote: > > > > > So I am guessing the Felix version is lucky there are > > > no gratuitous temporaries to be saved when this happens, > > > and the C code is unlucky and there are. > > > > > > Maybe someone who knows how the optimiser works can comment? > > > > One problem with departing from the ABI even on a local level > > like this is that it wipes out lots of tools that depend on > > ABI compliance for the entire call chain. I suspect the overall > > gain is too small to be worth this hit. > > If you make the function static then gcc can chose ABI-incompatible > calling conventions.
Yes, but to do so is reasonably hard because the choice of ABI now depends on usage, which is scoped to the whole translation unit. With a nested local function, it is scoped only to the parent function, and in the simple wrapper case there is only one call from the parent to the nested child, which doesn't have any performance constraints on it. So the only calls that matter when chosing the parameter passing model are the recursive ones inside the local wrapper. It is probably still 'hard' to choose a good parameter passing model even then. I used to write a lot of assembler code and always grappled with this. -- John Skaller <skaller at users dot sf dot net> Felix, successor to C++: http://felix.sf.net
