[Bug tree-optimization/106514] [12/13 Regression] ranger slowness in path query
rguenth at gcc dot gnu.org
gcc-bugzilla@gcc.gnu.org
Tue Dec 13 14:02:17 GMT 2022
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106514
--- Comment #11 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #10)
> Re-confirmed. =15 vs =30 goes from
>
> backwards jump threading : 0.58 ( 13%)
>
> to
>
> backwards jump threading : 7.00 ( 65%)
>
> so it still shows exponential behavior for CFGs like
>
> if (a)
> ...
> if (b)
> ...
> if (c)
> ...
>
> because we explore all paths through the CFG that fit in some size limit.
> I'm not sure if there's any low hanging fruit besides this (like if we
> properly avoid re-doing local computes when visiting blocks as part of a
> path multiple times).
We do have --param max-jump-thread-paths putting a hard limit on the number
of branches in a paths we explore, the default might just be a bit high
for this testcase.
More information about the Gcc-bugs
mailing list