[Cfp-interest 1712] WG14 IEEE 754-C binding meeting minutes 2020/07/22
Rajan Bhakta
rbhakta at us.ibm.com
Wed Jul 22 10:06:24 PDT 2020
Attendees: Rajan, Jim, David O., Fred, Mike, Ian, David H.
New agenda items:
None.
Carry over action items:
Jim: Check WANT macro issue in CFP 1581 - Done.
Last meeting action items:
All: Need ideas on how to convert Word to good PDF with links. Jim:
Add hex to annex draft. - Done.
Fred: Write paper saying ‘const’ missing in *payload. - Done.
Jim: Rewrite CFP 1643: Rnd ties away issue. - Done.
Jim+Fred: Re no prototype vs varargs prototype, find where in C where
float is promoted to double. - Done.
New action items:
Jim: Create a proposal based on CFP 1709 updates to Annex B with the
addition of N and M parameter descriptions.
Jim: For cfp3x-annex-20200706.pdf, Annex X, new example (page 33),
change the magic number 40 (which needs to be verified) to be a macro and
add in a descriptive definition of that macro.
Jim: For cfp3x-annex-20200706.pdf, Annex X, strtoencdecd* function
declarations (page 35), make the pointers into arrays to allow giving a
size for the array.
Fred: Submit the paper in CFP 1703 to WG14 as a CFP paper.
Jim: Submit CFP 1653's changes (appropriately colorized or changed to
highlight the changes) to WG14.
Jim: Put CFP 1634 (with the correction: 'prototype is extended' ->
'prototype is expanded') into TS3 as an annex.
Fred: Submit CFP 1704 to WG14 as a proposal from CFP.
Jim: Prepare a suggested change to C2X proposal based on CFP 1687
option 3.
Fred: Submit CFP 1702 to WG14 as a CFP paper.
David H: Come up with a sentence to add to footnote 295 (as per CFP
1697) to point out the possible numerical differences in output.
Next Meeting(s):
Wednesday, August 19th, 2020, 11:00 EST, 8:00 PST, 4PM UTC
ISO Zoom teleconference
Please notify the group if this time slot does not work.
C++ liaison
C/C++ compatibility for extensions floating types
See [Cfp-interest 1638, 1701] C++ floating-point work David Olsen
Difference between C++ and our TS in determination of types. C++
picks standard types whereas we pick the interchange types. It was an
issue for the tgmath functions before the rework (Ex. Functions that round
to narrower type).
In C++ you can have these new IEEE types and non-IEEE standard
types.
C2X Integration:
Latest C2X draft:
http://www.open-std.org/jtc1/sc22/wg14/www/docs/n2478.pdf
Part 1
Part 2
Part 3
Part 4ab
Part 5abcd
IEEE 754-2019 support
Action item details
Jim: Check WANT macro issue in [CFP 1581]
Some of the macros are missing in N2478 as well.
Rajan: Add in some text to describe what N and M are.
Resolve as per CFP 1709.
*AI*: Jim: Create a proposal based on CFP 1709 updates to Annex B
with the addition of N and M parameter descriptions.
All: Need ideas on how to convert Word to good PDF with links.
Adobe's online version does not work. The old Adobe Pro 9 works on
PC. Jim to try on Mac (has been ordered).
Jim: Add hex to annex draft [CFP 1684]
New example added (page 33).
Mike: Don't like seeing the "40" constant in the example.
Jim: Wasn't there a macro added for the maximum size Fred?
Fred: I don't think so.
Rajan: Do a "#define MAX_VALUE 40 // Max hex digits" and use that
in the declaration and strfromenc call.
Resolve as per CFP 1692.
*AI*: Jim: For cfp3x-annex-20200706.pdf, Annex X, new example (page
33), change the magic number 40 (review if this is big enough) to be a
macro and add in a descriptive definition of that macro.
*AI*: Jim: For cfp3x-annex-20200706.pdf, Annex X, strtoencdecd*
function declarations (page 35), make the pointers into arrays to allow
giving a size for the array.
Fred: Write paper saying ‘const’ missing in *payload [CFP 1655]
See CFP 1703.
*AI*: Fred to submit the paper in CFP 1703 to WG14 as a CFP paper.
Jim: Rewrite CFP 1643: Rnd ties away issue.
See CFP 1653.
*AI*: Jim: Submit CFP 1653's changes (appropriately colorized or
changed to highlight the changes) to WG14.
Jim+Fred: Re no prototype vs varargs prototype, find where in C where
float is promoted to double
See CFP 1688.
Fred: This is a change to the standard.
Jim: No, this is a change to the TS3 annex as per CFP 1634 (with
the correction: 'prototype is extended' -> 'prototype is expanded').
*AI*: Jim: Put CFP 1634 (with the correction: 'prototype is
extended' -> 'prototype is expanded') into TS3 as an annex.
Other issues
Constant expressions evaluated in translation environment [CFP 1579]
See CFP 1704.
*AI*: Fred: Submit CFP 1704 to WG14 as a proposal from CFP.
WANT macros for Annex F interfaces [CFP 1627]
See CFP 1687.
Option 1 is editorial, while 2 and 3 are changes.
Most seem to prefer option 3.
Suggested change to C2X.
*AI*: Jim: Prepare a suggested change to C2X proposal based on CFP
1687 option 3.
WANT macros for decimal and math functions [CFP 1657]
See CFP 1707.
Fred: Current version of Annex X has the
__STDC_WANT_IEC_60559_TYPES_EXT__ want macro in it.
Fred: Withdrawing my issue.
intmax_t removal [CFP 1617]
See CFP 1702.
Jim: The naming in the compoundn functions is different from other
functions like scalbn and scalbln to reflect the long argument.
Rajan: The change in return value from unspecified value to NaN may
cause questions.
*AI*: Fred: Submit CFP 1702 to WG14 as a CFP paper.
%a with formatting precision [CFP 1693]
See CFP 1700.
Ian: Should we be making a recommendation instead of a requirement?
Fred: Required to be correctly rounded.
Jim: Only within the possible representable set of values.
We could add a note that is essentially a warning.
David H: For portable code, they shouldn't get close to this case.
Jim: We could disallow a formatting precision of zero.
David H: That would be different from Decimal.
Fred: The problem is this applies to any precision lower than the
full precision needed for the value.
Jim: Different implementations can give different results. So not
portable.
Fred: Even with %a trailing zeros can be trimmed so not portable.
Ian: What is this useful for?
Jim: For showing what is in memory.
David H: If the conversion is less than full precision, the leading
digit has to be between 8-F.
Jim: Are we agreed we need a note about this?
David H: For precisions less than full, the choice of a leading
digit can differ between implementations.
Jim/David H: We should point out the numerical difference in the
output by expanding the footnote.
*AI*: David H: Come up with a sentence to add to footnote 295 (as
per CFP 1697) to point out the possible numerical differences in output.
Triple (1,0,0) [CFP 1675]
Discuss next meeting.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20200722/cb44ac7b/attachment-0001.htm>
More information about the Cfp-interest
mailing list