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

(fwd) 754r draft review minutes

from Jeff Kidder:


[I seem to not be on the list, could someone pass these on...]

IEEE 754r Meeting minutes July 21, 2004
Sun, Santa Clara

Attendees:      Affiliation
Dan Zuras       Self
Joe Darcy       Sun (phone)
Eric Schwartz   IBM  (phone)
David Hough     Sun
Jeff Kidder     Intel (minute taker)
Jim Thomas      HP
Peter Markstein HP
Jon Okada       HP
Michael Parks   Sun (host)
Mike Cowlishaw  IBM
Note taker: Jeff Kidder

Draft Review:
Page 14 - payload definition. Accepted.
What is the etymology of "payload"?

Pg 19 qNaN and sNaN -> qNaN (quiet), sNaN (signaling)

Pg 24 qNaN or sNaN - accepted

Pg 28 qNaN or sNaN - accepted

Pg 42 Summary of Operations - to be discussed in full session

Pg 58 Quiet Conformance Predicates - Should return bool.
Substantial discussion on the semantics of is754(x).
Removed the type. Included a purple note that we
need to figure out how to deal with the weakness of the
claim (only requires some radix and some single precision
to return True).

Pg 60 some discussion of redundant declets for min/max.
They should be made canonical.
"m_x" -> "c_x"
Add the term "arithmetic" before "operation"

Pg 78 NaNs
David proposed some new wording of the beginning of section N.
some discussion revisiting the sign of NaNs.

Pg 79
invalidNaN - "raise no exception" moved to the end.
Need to decide if this should raise invalid or not.
makesNaN - Need to figure out what to do with makeqNaN().

end of draft review

Discussion of dealing with payloads and int arguments.
Consider using strings.
Could be typedefed to be a struct or union or int, or ascii string.

Digit Strings:
Goal is to be fairly general -- don't say what the string encoding is.
We have a way to write constants. The standard provides ways to build
on them. Need to deal with the base of the strings.
Want to be able to create any payload.
Want to be able to do an extract and deposit which is the identity.
want to be able to treat them as integers in a particular range.
what would we want to do in Java? Perceived low thrust to weight.
Part I:
A canonical description for NaNs
in the conversion to strings and back.
Part II:
This is what the canonical string should be: ...

Next meetings:
David Hough is hosting in August.

754 | revision | FAQ | references | list archive