Math/trig functions vs -f68881

[or uunet!sli!rdh]

What is the meaning of "-f68881" to the Sun3 C compiler? I empirically
observe that it generates inline add/sub/mul/div floating-point
instuctions, but still generates calls to external math library/ies for
the math/trig functions (_fmod, _sin, etc.). What is it supposed to do? Is
there actually a specification somewhere in Sun's manual set -- I looked
in the Floating Point User's Guide, didn't find much help beyond a some-
what content-free 

  In-line code generation puts floating-point instructions in line with
  CPU instructions; such code generated in-line for specific hardware,
  ..., only runs on machines with the specific hardware...

Why doesn't -f68881 take advantage of the hardware sin/etc. functions?

You can email me (rdh @ sli.com   -or-   uunet!sli!rdh) directly, I'll
cull/summarize the replies.

Thankx in Advance!