[Cfp-interest] WG14 IEEE 754-C binding meeting minutes 2017/02/28
Rajan Bhakta
rbhakta at us.ibm.com
Tue Feb 28 10:35:31 PST 2017
2017/02/28, 12:00 EST:
Attendees: Rajan, Jim, David C., Mike, Marius, Fred, David H.
New agenda items:
None.
Last meeting action items:
All: Review the posted proposals (on CFP wiki with date 2016/09/01) by
next meeting as it is the last one before the Markham mailing. - Done.
Jim: DRS2: DDR7: Look to see if there is a simpler way of showing the
change. - Done (2017/02/20 email).
Rajan: DRS2: DDR9: Add code examples for implementers to test out and
send it out via email. - Done.
Jim: DRS2: DDR2: Write a paper with this text
(DECIMAL_DIG_obsolescence.pdf) for WG14. - Done.
Jim: Make SC22WG14.14586 a proposal for C2X. - Done.
Jim: Respond to Joseph's comments on SC22WG14.14586, SC22WG14.14587
via WG14 reflector. - Done.
New action items:
Rajan: Submit DRS2:DDR9 example as a paper for a supplement to the STC
in the DR and change all the %A to %a (ensuring it conforms to the
specification for %a).
Jim: Create a DR against part 2 to specify the capitalization effects
of %a.
Jim: DRs2:DDR2: DECIMAL_DIG: Have DECIMAL_DIG to be the largest
supported floating type as a new replacement Suggested TC for DRs2:DDR2.
Jim: Add in the encoding *_DIG macros (as per
http://wiki.edg.com/pub/CFP/WebHome/characteristics_for_non-arith_formats.pdf
) and obsolescing DECIMAL_DIG as a proposal to WG14 as a new enhancement.
Jim: Include Type-generic macros for functions that round result to
narrower type as part of the third set of DR's.
Fred: Propose adding in #pragma STDC FENV_ROUND DEFAULT as a means to
set the static rounding mode to the default rounding mode.
Next Meeting:
Tuesday March 21st, 2017, 12:00 EDT, 9:00 PST
Same teleconference number.
Discussion:
IEEE 754 revision:
Next meeting will be the last one for any substantive changes.
min/max: Not reflective of NaNs. Can have them as most important or
least important. Recommendations so won't break anyone.
We will need to name it too somehow.
C++ liaison:
Ian not present.
What should be proposed for the C standard (C2X):
Part 3 is one proposal. Extended, interchange, and support for
interchange formats. Can split it up if needed.
Keep it as one proposal, but mention it can be split up if needed.
Part 4 is two proposals: transcendental functions and reduction
functions.
Keep it as two proposals.
Jim: Is 754's thinking changing with regards to the reduction
functions?
David H: Been talked about due to vectorization but leaving it
as is since it is too big an item to handle right now.
Jim: Should we lower the priority or not push for this due to
the expected future changes?
David H: How stringent things are about exceptions is likely to
change. The basic functionality should be the same and shouldn't contract
but may be enhanced.
Part 5 is 4 proposals: Expression evaluation, optimization,
reproducibility, alternate exception handling
Leave it as is.
Jim: The alternate exception handling has a natural split on ones
that change control flow.
The ones that don't change control flow are fairly
straightforward, so may be easier to pass since less invention.
We do have a reputation of being flexible so it should not be a
problem about doing parts and not others.
We have draft documents for everything and prior art.
2017/02/02 email for proposal for C2X for FE_TONEARESTFROMZERO
rounding direction (
http://wiki.edg.com/pub/CFP/WebHome/C2x_proposal_-_FE_TONEARESTFROMZERO.pdf
)
Go ahead with the proposal.
DRs:
Discuss suggestion in Jim’s Feb 20 email “simplify DRs2:DDR7”
Jim: Should be non-substantive change in part 2, but not sure.
This means the part 3 tgmath will rely on the usual arithmetic
conversions.
Treat this as a new paper for a new simpler suggested TC for DDR7
so the committee has a choice on which TC to take.
Follow-up on DRs2:DDR9 a-style formatting not IEC 60559 conformant.
See code examples in Rajan’s Jan 24 email “Examples to add for
Defect report set 2, Draft defect report 9”.
Fred: Change return 55 to return 0.
*Rajan: Submit this as a paper for a supplement to the TC in the
DR.
Discuss Jim’s Jan 24 email “RE: Examples to add for Defect report
set 2, Draft defect report 9” about effect of %a vs %A for decimal.
*Jim: Should we create a DR against part 2 to specify the
capitalization effects?
Yes.
Rajan: Check the specification now to make sure the example
conforms to what is currently specified.
Discuss DRs2:DDR2. See Jim’s Feb 1 email “AI: DECIMAL_DIG
obsolescence” and Joseph Myers’s Feb 20 email “Re: (SC22WG14.14602) N2108
and 18661-3 definition of DECIMAL_DIG”. See also Ian’s Jan 20 email “Re:
AI: proposal to obsolesce DECIMAL_DIG”.
Due to the encoding functions, we need a size to allow buffers for
non-arithmetic types.
Even with
http://wiki.edg.com/pub/CFP/WebHome/characteristics_for_non-arith_formats.pdf
we need to make DECIMAL_DIG equivalent to LDBL_DECIMAL_DIG or maybe just
make it the largest supported floating type (Ian's suggestion).
*Jim: DECIMAL_DIG: Have DECIMAL_DIG to be the largest supported
floating type as a new replacement Suggested TC for DRs2:DDR2.
*Jim: Add in the encoding *_DIG macros (as per
http://wiki.edg.com/pub/CFP/WebHome/characteristics_for_non-arith_formats.pdf
) and obsolescing DECIMAL_DIG as a proposal to WG14 as a new enhancement.
Discuss DRs3:DDR2 Type-generic macros for functions that round
result to narrower type.
See posted DDRs3-20170121.pdf.
*Jim: Include Type-generic macros for functions that round result
to narrower type as part of the third set of DR's.
Discuss suggestion in Fred’s Feb 22 email “Part 1: FENV_ROUND
DEFAULT”.
Sounds good for non IEEE floating point types.
*Fred: Propose adding in #pragma STDC FENV_ROUND DEFAULT as a
means to set the static rounding mode to the default rounding mode.
Other:
None.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20170228/70592c3c/attachment.html
More information about the Cfp-interest
mailing list