[Cfp-interest 1320] WG14 IEEE 754-C binding meeting minutes 2019/05/21
Rajan Bhakta
rbhakta at us.ibm.com
Tue May 21 10:13:37 PDT 2019
2019/05/21: 10:00 CST:
Attendees: Rajan, Jim, Mike, Blaine, David H,
New agenda items:
None.
Carry over action items:
All: Review the rationale for part 5 a, b, c proposal. - Carry over.
Fred: Create papers for the SNAN initialization and unary + operation
as CFP papers (CFP 1249, 1253, 1247, 1250) for future submission to WG14.
- Carry over.
Last meeting action items:
Ian: Forward message from Hubert about FLT_EVAL_METHOD, etc. - Done.
Jim: Post the IEEE 2019 draft for the CFP group on the wiki. - Done
(see 4/25 note).
Jim: Ensure that the quantum exponents table defines dN sufficiently
in C2X. - Carry over (review next draft).
Jim: Get an N number for CFP1277 and submit it. - Done.
Jim: Get an N number for CFP1282 as a proposal and submit it. - Done
(CFP review).
New action items:
Jim: Investigate creating our own CFP compendium.
Fred: Give a new version of the SNAN initialization paper (as per
CFP1316).
Jim: Point out to Jens that we're using two spellings for analog in
the current C2X draft.
Jim: Look into the commas needed in the
why_no_wide_string_strfrom_functions document, then get a document number
and submit it.
Jim: Keep Rounding of negated constants on the agenda to discuss for
next meeting.
Jim: Keep fesetexcept on the agenda to discuss for next meeting when
Fred is present.
Next Meeting(s):
Wednesday, June 26th, 2019, 11:00 EST, 8:00 PST, 4PM UTC
Same teleconference number.
Please notify the group if this time slot does not work (note the new
day).
Discussion:
754 revision:
See CFP 1317.
Draft 50 is the technical content that will go forward.
C++ Liaison:
None.
WG14 meeting (April 29th-May 3rd):
CR process going forward, with Blaine Garst.
See CFP 1318 (Blaine’s email forwarded by Jim).
Jim: The review cycle is not too short. It is good. Also us taking
over the the compendium is reasonable. A lot of what we had is in C2X, so
any changes to parts 1-4a will go through the C2X process. Our compendium
will be very small now. I liked the meeting history in the original
compendium. I will keep that.
Jim: Currently CR's come with some suggested change, rather than
questions.
Blaine: We're not clear yet as a committee. I don't think the
compendium should carry normative content forward. The editors can make
the changes directly. Still need to work out the process.
Jim: I currently put changes into my editors draft as soon as we
have a CR/DR. Nothing goes out until it is closed.
Mike: Often have to go back and make changes to text people want
in immediately.
Blaine: Having the committee agree to the words makes this much
easier.
*AI*: Jim: Investigate creating our own CFP compendium.
See CFP 1309.
N2323 will have a new N document (N2379) that doesn't talk about
operations which was what was voted in as direction.
N2326: Jim: I think there is a blurring of value and
representation. Instead of normalized it should be normal number. It has a
value of some normalized form in the model. I think this needs some more
thought.
C2X integration:
Part 1 – Integration completed.
Part 2 – Integration mostly done.
Part 3 – Expecting N2342 is sufficient for the WG14 editors.
Part 4a – We need to work on something to give to the WG14 editors
for integration into C2X.
Part 5a,b,c,d – Considering new proposals for a,b,c
Action item details:
Fred: Create papers for the SNAN initialization and unary + operation
as CFP papers (CFP 1249, 1253, 1247, 1250) for future submission to WG14.
See Tydeman’s CFP 1290. Jim’s CFP 1316.
*AI*: Fred: Give a new version of the SNAN initialization paper (as
per CFP1316).
All: Review the rationale for part 5 a, b, c proposal.
http://wiki.edg.com/pub/CFP/WebHome/n2120.pdf,
http://wiki.edg.com/pub/CFP/WebHome/n2121.pdf,
http://wiki.edg.com/pub/CFP/WebHome/n2122.pdf
Ian: Forward message from Hubert about FLT_EVAL_METHOD, etc. See Ian’s
CFP 1287.
Jim: Seems C++ takes C FLT_EVAL_METHOD.
Jim: Ensure that the quantum exponents table defines dN sufficiently
in C2X.
Jim: Jens wanted to use or refer to the macros in tgmath, and I
wanted to avoid referring to tgmath. This will be an item for the next
meeting.
Jim: Get an N number for CFP1277 and submit it.
http://wiki.edg.com/pub/CFP/WebHome/n2384.pdf
Jim: Get an N number for CFP1282 as a proposal and submit it (possibly
after CFP review?).
http://wiki.edg.com/pub/CFP/WebHome/C2x_proposal_-
_why_no_wide_string_strfrom_functions.pdf
Mike: Slight editorial comment: Should avoid spellings that have one
side identical to the other.
*AI*: Point out to Jens that we're using two spellings for analog in
the current C2X draft.
Mike: Should be a comma after the etc. at the end too.
*AI*: Jim: Look into the commas needed in the
why_no_wide_string_strfrom_functions document.
Other issues
Rounding of negated floating-point constants under FENV_ROUND pragma.
See Jim’s CFP 1314 and Mike’s 5/14 reply.
Jim: Same issue if you add more digits to the constant.
Mike: Not really.
Jim: In C the negation is an operator, not part of the constant.
Matters for +inf or -inf rounding, which is different from what
strtod or scarf would do.
Mike: They would have to know but programmers do have a choice with
this.
Jim: Could be done with parenthesis around the constant too.
Mike: It is an issue for unary + as well.
Jim: It shouldn't be. This is only for unary -.
Jim: C doesn't have any way of representing negative values as a
constant.
*AI*: Jim: Keep Rounding of negated constants on the agenda to
discuss for next meeting.
fesetexcept and optional inexact
See CFP email thread “fesetexcept() and optional inexact”
*AI*: Jim: Keep fesetexcept on the agenda to discuss for next
meeting when Fred is present.
Fred’s WG 14 papers:
See WG14 email thread “N2380: printf of NaN()”
Jim: The sign of the NaN is also not portable across
implementations.
Mike: This sounds like the most non-problem I can think of.
Existing portable code already has to deal with this along with
all other implementation defined behavior.
Jim: 60559 has payloads being implementation specific. No goal of
making them portable, the opposite in fact.
Mike/Rajan: This can be done portably now with output to a sprintf
buffer and processing the NaN string.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20190521/0264cbec/attachment-0001.html
More information about the Cfp-interest
mailing list