https://gcc.gnu.org/bugzilla/show_bug.cgi?id=38518
--- Comment #15 from David Binderman <dcb314 at hotmail dot com> ---
(In reply to David Binderman from comment #12)
> With recent gcc:
>
> no -O flag: 2 seconds
>
> -O2: 24 seconds
>
> -O3: 247 seconds
Those numbers for a version of trunk gcc with lots of checking.
Without checking, I get gcc trunk with -O3 in 4 mins 56 seconds
and release gcc 7.1.1 in 3 minutes 38 seconds, so it still looks like it's
going backwards.
From -ftime-report on trunk gcc with -O3:
$ grep "([0-9][0-9]\%) usr" report.out
df live&initialized regs: 45.04 (16%) usr 0.06 ( 2%) sys 45.74 (15%) wall
0 kB ( 0%) ggc
load CSE after reload : 174.76 (60%) usr 0.07 ( 2%) sys 176.23 (60%) wall
10 kB ( 0%) ggc
$
So I think that confirms what Richard is seeing.
