[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Issue: FLOAT-UNDERFLOW (version 2)

    Date: Wed, 24 May 89 08:15:24 PDT
    From: Jon L White <jonl@lucid.com>

    Lucid 3.0 and later has LEAST-{NEGATIVE,POSITIVE}-NORMALIZED-<mumble>-FLOAT,
    and in fact copied the names from Symbolics.  These, and the prescription 
    that LEAST-{NEGATIVE,POSITIVE}-<mumble>-FLOAT be denormalized in 
    implementatons which support it, seem very non-controversial to me.

Thanks, I'll add that to the current practice section in the next version.

    But WITHOUT-FLOATING-UNDERFLOW-TRAPS is too limited.  The topic needs
    more thought, because much more than "underflow" should be considered.
    Lucid 3.0 and later has WITH-FLOATING-POINT-TRAPS, which takes two
    lists of condition names relevant to floating point operations and
    selectively enables or disables them (one list for "enablements", and
    one for "disablements").  And I wouldn't like to bet on our being
    able to achieve consensus on this design over the next few weeks,
    even though I agree that it is an important topic.

The inability to converge on a design for such an elaborate feature is
precisely the reason for proposing the very simple feature.  Also note
that underflow is the _only_ exception that some applications have a
very strong need to enable while at the same time other applications
have a very strong need to disable.  Thus just the simple feature gets
us most of the way towards perfection.  By the way if you want to write
up a proposal for the Lucid WITH-FLOATING-POINT-TRAPS so we can discuss
it, that would be fine with me.  But I will be very disappointed if the
end result is that we can't agree on it and do nothing at all.