[cfp-interest 3660] JTC1/SC22/WG14/CFP 2025/10/15 Meeting notes
Jerome Coonen
jcoonen at gmail.com
Wed Oct 15 20:56:05 PDT 2025
2025/10/15: 8:00 AM PDT/3:00 PM UTC
[Please submit proposed changes to these minutes to Jerome or to the group.
Revision changes appear at the bottom.]
*Attendees*
Jim Thomas, Rajan Bhakta, Jerome Coonen, Damian McGuckin, David Hough,
Tue Ly, Joshua Cranmer, Eskil Steenberg Hald
*Updated agenda plus new items*
https://wiki.edg.com/pub/CFP/WebHome/CFP meeting
agenda-20251015-update.pdf – These minutes should be read alongside the
agenda, with its many reference links.
*Previous meeting notes*
https://wiki.edg.com/pub/CFP/WebHome/n3725.pdf
*Study group logistics*
Next meeting: 29 October 2025, 8:00 AM PDT/3:00 PM UTC
Final 2025 meeting 12 November 2025, 8:00 AM PDT/3:00 PM UTC
ISO Zoom teleconference
Please notify the group if this time slot does not work.
*C documents*
The latest C2Y draft is N3685 Sept. 2025
https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3685.pdf
C23 has been published ISO/IEC 9899, available for purchase.
https://www.iso.org/standard/82075.html
*IEEE 754 liaison*
David: Making gradual procedural progress.
*C++ liaison*
Joshua: Next meeting first week of Nov. Two papers on floating point
topics: defining fastmath semantics is hard and reproducibility.
*WG14 update*
Rajan: Next meetings 2-6 Feb 2026.
*TS-4 and TS-5 revisions*
None.
*News*
Jerome: (a) The 754 chair announced an invitation-only 40th anniversary
event for 754 on 1 Nov in Berkeley, at the site of the IEEE Milestone
plaque dedicated May 2023. (b) The IEEE Milestone plaque for the Intel 8087
Math Coprocessor was finally dedicated on Haifa, Israel. It was delayed
over a year by the troubles there. (c) The Computer History Museum has just
posted the oral history of Ravi Nave's work as leader of the silicon design
team for the 8087. Jerome will send a note to the 754 and CFP groups.
*Carryover action items from last meeting *
None
*Action items from last meeting *
Jim & Rajan: Send email to WG14 seeking direction for how to deal with
translation time evaluation of constant expressions that does not lead to
possible undefined behavior.
Done
Jim & Rajan: Craft note to Alex, CC to WG14, re. n3584, removing
imaginary I.
Done
Jerome: Draft a proposal re. the language of functions returning the
principal value.
Done
Jim & Jerome: Look further into the language around when errors "occur"
in the sense of C2Y and how the setting of errno fits in. Draft a proposal,
if appropriate.
In progress, carry over
*Discussion of issues*
Floating expression evaluated in the translation environment
Rajan: Action was to remove undefined behavior. Remove the
occurrence of "shall" that allows UB, noting that the translation
environment knows the bounds of range and precision of the execution
environment. Action to submit.
Removing imaginary I
Rajan: Alex has not responded to note but the topic will arise in
the next WG14 meeting. No further action now.
TOWARDS and missing quotes issues
Jim: These are addressed in the latest C2Y draft. Closed.
Error handling in TS-5
Jim: Joseph has updated WG14 Issue 1014 to record the proposed
correction and its acceptance (when TS-5 is ever updated). No further
action.
Output bounds for math functions
Jerome: Jim dug into this after previous meeting, doing the lions
share of the action. This brings the language of C2Y in line with 60559 and
removes a conflict with correctly-rounded implementations. Action to submit.
Special cases of cpow
Jim: Paul Z. sent email about this, requesting guidance on handling
of special cases. Jim and Paul have corresponded. This is a request for
information, not a request for action.
Preferred quantum exponent for nextafterdN
Jim: Fred brought up the situation of quantum exponents for
“returns y when x equals y". Jim proposes language along the line "returns
the numerical value of y when x equals y” with reference to Table 5.2 for
the preferred quantum exponent. Action.
Annex F changes discussion paper
Damian: Reviews several specific items. Removing the comma
introducing compound cases "for x < C n y > 0" would be consistent with
other C2Y usage. Use the singular form "integer n". A big issue is the
choice between single-ended and double-ended bounds for domains.
Jim: The standard doesn't use ranges but uses constraints like "0 <
x".
Damian: Looks at atan2() cases as an example. Is "finite y > 0"
better than "0 < y < +inf"? Is "y > 0" better than "0 < y <= inf", given
that inf is assumed to be part of the augmented reals? Then compare cases
from tanpi(), with language like "evn n > 0". Here the modifier "even" is
all but required.
Ly: Proposes calling out all the inf cases and having "x" always be
finite.
Rajan et al.: That is too large a change.
Damian & Jerome: Annex G is complicated enough to require explicit
statements of bounds. But does Annex F have to be expressed just like G?
The audience for F is wider and may benefit from the simpler, more concise
language that can work for F but not G.
Damian: Will do two versions, with "finite" and with bounds.
F.3 #7 quiet/signaling NaNs
Jim: Annex F specifies optional support for signaling NaNs, but the
specification doesn’t cover the signbit macro. The signbit macro is one of
a family of operations not sensitive to signaling NaNs.
Jerome: Take action to find the macros and functions that handle all
nans silently.
Encode/decode in TS-4
Damian: Forwarded Michel's message re. 60559 because it's already
in Annex H.
Jim: No action.
Overflow, non-default rounding, errno
Action carried over.
F.10.9.1 - nextafter(x, y)
Damian & Jim: There is an issue of the language of the return value
of a function. Using simple "result" can be ambiguous.
Eskil: "Function value" might be misinterpreted as a pointer to the
function.
Jim: Take action to write a note about what to do.
strtodN clarification
Jim: Fred raised the issue of clarifying the type the conversion is
rounded to.
Jerome et al.: Does not seem to be ambiguous.
Rajan: Proposes adding to low-priority list. Action for Jim.
*Other issues*
None.
*Adjournment*
10:02 AM PDT
-----------------------------------------------
*Action items to be carried over*
Jim & Jerome: Look further into the language around when errors
"occur" in the sense of C2Y and how the setting of errno fits in. Draft a
proposal, if appropriate.
*New action items*
Rajan: Submit the proposal regarding translation time expression
evaluation.
Jim: Submit the proposal regarding output bounds for math functions.
Jim: Draft a proposal for the language of quantum exponent
in nextafterdN.
Damian: Draft Annex F change proposal in two flavors, with modifiers
and with double-ended bounds.
Jerome: Follow up on 3618 to see whether snan issues apply to more than
signbit, and what to do about it.
Jim: Write a note about the locution of "result" vs. "result value" vs.
"return value" in some problematical functions where the language is not
best today.
Jim: Add strtodN to list of low-priority issues.
*Discussion issues to be carried over*
F.10.1#14
Rounding modes, terminology
canonicalize() in AnnexF - F.10.9.7
hypot(x, y) - C2Y F.10.5.4
hypot(x, NAN) in Annex F
Magnitude or Absolute Value or |x| or ....
Preferred Style in Annex F - fromfp/fromfpx or nearbyint/rint
Consistency in paragraphs using NaNs
The nan functions
Occurs vs shall occur for math errors
Floating-point function or function returning a floating-point result
SNAN macro location change
*Signoff*
Respectfully submitted.
-Jerome Coonen
650.996.4738
jcoonen at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20251015/ba39e2b7/attachment.htm>
More information about the cfp-interest
mailing list