Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

As simple as it is now, I am still against motion 24.03...



	Folks,

	The motion is now, & I quote:

	The Motion: Every IEEE 1788 compliant system SHALL provide
	the four basic arithmetic operations addition, subtraction,
	multiplication, and division with rounding downwards and
	upwards. Type conversions with directed roundings SHALL
	also be provided.

	(Emphasis mine.)

	I agree with Prof Kulisch that these operations are a
	desirable way to bring today's machines into an era in which
	hardware interval arithmetic is no more unusual than hardware
	floating-point arithmetic.

	I even agree that, were I to advise a designer of a new
	computer chip TODAY, I would advise that statically
	defined directed rounding arithmetic be made as easy &
	fast as round-to-nearest arithmetic.

	But I disagree that we should mandate that step with a
	shall.

	Indeed, we are not publishing 1788 today.  And when we do
	publish it we may have moved into an era where such an
	approach to hardware implementation of interval arithmetic
	is actually counter productive to fast interval hardware.

	(I am thinking of hardware which implements entire interval
	operations, including management of decorations.  But I may
	be wrong.  There may be other approaches none of us have
	thought of yet that serve us better.  In such a situation,
	mandating directed rounding instructions that will never be
	used presents an unneeded barrier to acceptance of 1788.)

	Therefore, I am against a mandate (shall) on the grounds
	that it might hurt us as easily as help us.

	If those two shalls were turned into shoulds I would vote
	for the motion.  But Prof Kulisch knows as well as I do
	that the shoulds might as well not even be there.

	So I will vote against motion 24.03 & urge others to do so
	as well.


				   Dan