[Bug tree-optimization/108500] [11/12 Regression] -O -finline-small-functions results in "internal compiler error: Segmentation fault" on a very large program (700k function calls)

[rguenth at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108500

--- Comment #15 from Richard Biener <rguenth at gcc dot gnu.org> ---
To not look at "nothing" (after successful SRA it should indeed become almost
nothing) I've added a store to a volatile 'x' global variable to the end
of main:

...
  s2 = f(s1,s2);
  x = s2;
  return 0;
}

not using main but a function returning a struct would probably work as well.
Otherwise DCE/DSE will remove all code.  Doing that reveals that RTL DSE
(more specifically rtx_equal_for_cselib_1) is very slow (via
dse_step1 calling cselib_process_insn).

I do wonder if you can share the "real" testcase?  It doesn't need to
be able to link, preprocessed source would be enough.