754R meeting December 15
The 754r meeting announced by Dan Zuras will take place on Thursday
December 15 at Sun Microsystems at Menlo Park.
Agenda includes a technical presentation by Paul Zimmermann
and other items as previously announced.
There will also be a non-754R meeting on December 14 to which interested
754R members are also invited.
Agenda consists of a technical presentation by Paul Zimmermann.
Paul Zimmermann will be making technical presentations at both meetings
as indicated below.
These meetings will take place at Sun Microsystems Menlo Park
starting at 1pm and adjourning between 5-7pm.
!! For all these meetings, it is very helpful to register your intention to
!! attend in person,
!! as far in advance as possible, to expedite your processing by Sun
!! security. Please register by emailing me: 754r@xxxxxxxxxxx
On meeting day around 1:00pm, report to the lobby of
Sun Menlo Park Building 12, at 12 Network Circle. (Note different lobby
Sun's Menlo Park campus is located at the east end
of Willow Road (exit from 101) and near the west end of the Dumbarton
Bridge (hwy 84, exit from 880).
At 1pm I or another Sun employee
will be at the MPK12 lobby to meet visitors. If you arrive at other
times, the receptionist will have to call the meeting room to arrange for an
escort, or call my cell phone 408-202-1102.
Persons planning to participate by phone should call
and use participant access code
DATE : 15-Dec-2005
TIME : 12:30 pm to 5:30 pm PST
BLDG : MPK16
ROOM NAME : CAPITOLA
ROOM NUMBER/FLR : 3429/3
EXTENSION : Phone : 89462
Title: Why transcendentals and arbitrary precision?
Abstract: The IEEE 754 standard requires correct rounding for the four
arithmetic operations and the square root only. The 754R revised standard
will (or will not) require correct rounding for transcendentals. In this talk,
we show why we should (and how we could) normalize transcendentals. This is
joint work with David Defour, Guillaume Hanrot, Vincent Lefèvre, Jean-Michel
Muller and Nathalie Revol, which was already presented to the committee by
In the second part of the talk, we explain why it is useful to extend the IEEE
754 philosophy to arbitrary precision, and to develop and maintain
corresponding software implementations.
DATE : 14-Dec-2005
BLDG : MPK16
ROOM NAME : MARTINS BEACH
ROOM NUMBER/FLR : 3436/3
EXTENSION : Phone : 89460
Title: Error bounds on complex floating-point multiplication
Abstract: Given floating-point arithmetic with t-digit base-B significands
in which all arithmetic operations are performed as if calculated to infinite
precision and rounded to a nearest representable value, we prove that
the product of complex values z0 and z1 can be computed with maximum absolute
error 1/2 sqrt(5) |z0 z1| B^(1-t). In particular, this provides relative
error bounds of 2^(-24) sqrt(5) and 2^(-53) sqrt(5) for IEEE 754 single and
double precision arithmetic respectively, provided that overflow, underflow,
and denormals do not occur.
We also provide the numerical worst cases for IEEE 754 single and double
This is joint work with Richard Brent and Colin Percival.