https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127587

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
I thought as we're summing small integers, the result would be exact.  Of
course -ffast-math allows divergence, still from a QOI perspective I don't
expect any.

So why does a difference appear?

Reply via email to