[Bug rtl-optimization/70625] [4.9/5 Regression] Memory exhaustion when building specific snippet at -O2

["ktkachov at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70625

--- Comment #3 from ktkachov at gcc dot gnu.org ---
(In reply to Arseny Solokha from comment #2)
> (In reply to ktkachov from comment #1)
> > I see the PR70623 ICE with 4.9 and 5 at -O2 on arm-none-eabi.
> > Do you have any other relevant gcc configure options that trigger this?
> 
> Talking about this particular PR, I can't even reproduce it w/ all the
> targets I have: it works well w/o ICE for powerpc-e300c3-linux-gnu, for
> example.
> 
> As of PR70623, for me it started triggering ICE somewhere between
> 6.0.0-alpha20160110 and 6.0.0-alpha20160403 for both
> powerpc-e500v2-linus-gnuspe and x86_64-pc-linux-gnu. I failed to reproduce
> it w/ 4.9.3 and 5.3.0 configured for all four targets mentioned by me here.
> Given your comment, I'll certainly try it in the days ahead w/ the flavor of
> arm-none-eabi that I have.
> 
> In case I've got you wrong, what configure options do you want me to make
> specific?

I tried compiling it for an arm-none-eabi target and I get the ICE rather than
seeing the memory-hog behaviour.
By configure options I mean the output of gcc -v.
For arm targets, the --with-arch,--with-cpu,--with-fpu,--with-float-abi
configure switches can result in different compiler paths being taken, which
may expose or hide the bug.