[Bug target/86425] Spec 2006 soplex seems to be slower on PowerPC using -ffast-math than without -ffast-math

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86425

Martin Jambor <jamborm at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hubicka at gcc dot gnu.org,
                   |                            |jamborm at gcc dot gnu.org

--- Comment #1 from Martin Jambor <jamborm at gcc dot gnu.org> ---
On AMD Zen CPUs at least, we found that the number of iterations executed by
the hottest loop is considerably higher with -ffast-math (just patch the
benchmark and see for yourself).  The reason is that the benchmark iterates
until some error is smaller than a given threshold and with -ffast-math we
simply get there after more computation.