Do you need dynamic rounding modes?

[David Keaton]

     In the description of the DEC Alpha that was just posted to
comp.arch, there is an interesting comment on the IEEE flags.

>Rather then keeping global state bits for arithmetic trap enables and
>rounding mode, these enable and mode bits are encoded in the function field
>of each instruction.

     Apparently, then, rounding direction becomes a static mode
rather than a dynamic one.  Is this evil, or can you live with it?
If you can live with it, is it alright for compilers to assume
a static rounding direction even on hardware that allows it to be
dynamic (assuming this allows some additional optimization somewhere)?

					David Keaton
					Cray Computer Corp.
					dmkacraycos.com