[Bug target/124434] long double exceptions are 56x slower

[gnu at skewray dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124434

--- Comment #3 from Brian M. Sutin <gnu at skewray dot com> ---
The issue only appears on -O0: no options. It goes away with -O1, -O2, -O3. I
never run denormals, only good numbers (sum converges to 2.0), INFINITY, or
NAN. (I should check denorm...)

The dependence on optimization is what made me think this was a gcc issue.