https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125231

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
> Here is the function that calculates power...
> https://pro.spec.org/private/osg/cpu/cpuv8/analysis/coverage/kit87/729.abc_r/$
>
> ...and returns the value to get printed out at line 346 in the same
> file. This coverage is from an older kit, but I believe it is still
> valid here. (Also, I am working on posting full coverage of 1.0.1).

Thanks, there's indeed a reduction with float precision involved 
but we do not vectorize any of the loops involved, so the
error is likely in the data flowing into the power computation ...

Reply via email to