What is acceptable for -ffast-math? A numerical viewpoint

[dewar@gnat.com]

<<So, concluding: if you have programs that run for several days, you'd be
happy if you could cut that time by some hours using optimizations that
>>

The idea that these optimizations can save time of this magnitude is without
evidentiary foundation. Sure, if an optimization can save a lot of time, we
look at it carefully. But if not, then it is not worth the time to implement
it, or the uncertainty and complexity that it introduces, even if the result
is identical. In the case where the result is also degraded, the burden is
greater.