Fourth Draft "Unsafe fp optimizations" project description.

[dewar@gnat.com]

<<I can send you a copy if Intel has cleaned up its web site. I don't have
the Ada requirements but I can't believe that anybody could seriously ask
for more accuracy (you never know until you read, however). The behaviour
on out-of-range inputs is another matter, although I believe that anybody
calling direct trigonometric functions (sin, cos and especially tan) with
large arguments should be aware of the risks he/she is taking.
>>

Well Ada has very specific requirements (see ARM Annex G, even 
for large arguments).

The issue is not accuracy in general, but the requirement that there not
be a single case in which the result from the hardware is outside ther
required model interval specified in the ARM. I looked at the Intel
documentation quite a while ago, and it was statistical in nature,
and did not have any absolute guarantees. That's what worried me.