> I developed this patch while working on the dwarf2 pass series. > This was before I bypassed the entire problem by removing the > !deep branch prediction paths. > > Ideally, we'd do this generically from gimple. Less ideally, > but still better, is to always emit rtl, and support that in > the middle end without so many hacks in the back end.
Being able to emit variadic thunks through the standard channels would be cool. I was thinking about this to make cgraph code not having to special case the thunks, but eventually gave up concluding that it is very nnatural to actually repesent thunk call in gimple. The patch looks fine to me, especially because we will need fewer hacks for x32.. Honza