[Bug tree-optimization/128517] [14/15/16/17 Regression] wrong code at -O3 -march=x86-64-v3: niter drops the assumptions of all exits of a vectorized early-break loop since r14-6822

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128517

--- Comment #4 from Robin Dapp <rdapp at gcc dot gnu.org> ---
> it would be interesting what changed this.  ISTR we also wanted to get
> away from using LOOP_C_FINITE - Robin?  The simplest fix might be to
> simply not set LOOP_C_FINITE for early-break loops.

Yeah, the patch exists locally but is still rough and isn't well tested yet.  I
guess it qualifies as bug fix, i.e. for stage 3?  If so, I'll post it at some
point but if it's a stage 1 thing, I'd rather postpone.