[Bug rtl-optimization/64164] [4.9/5 Regression] one more stack slot used due to one less inlining level

[aoliva at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64164

--- Comment #20 from Alexandre Oliva <aoliva at gcc dot gnu.org> ---
(In reply to Alexandre Oliva from comment #18)
> No, we don't even *consider* the coalescing performed in the -DOPT case,
> because, as noted in comment 13, the SSA names ended up with different base
> names, because copyrename wouldn't give them the same base name.
> 
> (In reply to Richard Biener from comment #17)
> > To add to all this - IMHO copyrename should go
> 
> That's fine with me.
> 
> > Yes, out-of-SSA coalescing could be changed to allow coalescing of SSA names
> > with a user-DECL and anonymous SSA names (or SSA names with a DECL_IGNORED_P
> > decl).  But that will make the conflict graph much larger(?).

Not only that: unless we allow also coalescing between different SSA names in
out-of-ssa, we'll lose optimizations that copyrename would have enabled
out-of-SSA to make, because the removed copyrename won't have assigned the same
base name to otherwise unrelated SSA names that out-of-SSA won't even consider
for coalescing.  Now *this* might make the conflict graph explode in size.