(j3.2006) Liaison to IFIP WG 2.5
Wed Aug 22 17:13:57 EDT 2007
On Wed, 2007-08-22 at 15:51 -0500, Bill Long wrote:
> Well, if the trade-off between a couple of multiplies and a 9-way
> branch nest is not obvious to Kulisch, it is no wonder that computer
> people don't taking him seriously.
Kulisch understands this. What he outlines in his book is how a
floating-point unit could do the calculations at almost the same speed
as point arithmetic, with only two arithmetic units instead of four, and
about three more gate delays. This is very different from killing the
pipeline, or requiring four arithmetic units plus two four-way compare
The software solution of doing four multiplies might well be better than
two multiplies within each branch of a nine-way where/elsewhere block,
but of course it only helps at all for goodly-size array computations --
big enough ones to cover up the killing of the pipeline the occurs when
changing the rounding mode.
Van Snyder | What fraction of Americans believe
Van.Snyder at jpl.nasa.gov | Wrestling is real and NASA is fake?
Any alleged opinions are my own and have not been approved or
disapproved by JPL, CalTech, NASA, the President, or anybody else.
More information about the J3