[Cfp-interest] CFP Study Group Teleconference - Thursday, February 14

[Jim Thomas]

Fred,

Thanks for the quick look. See responses below. I intend to post an update tonight.

-Jim

On Feb 12, 2013, at 9:20 AM, "Fred J. Tydeman" <tydeman at tybor.com> wrote:

> On Mon, 11 Feb 2013 22:11:37 -0800 Jim Thomas wrote:
>> 
>> I posted a rough first version of Part 3 (Interface 
>> and extended types) on the wiki, for review on 
>> Thursday, and plan to post an agenda tomorrow.
> 
> Very good first draft.
> 
> A few comments.
> 
> PDF page 8, lines 9 and 10: Part 1 and Part 2 need more.  Perhaps: Of
> this technical specification.

Right. These are just placeholders.

> 
> PDF page 9, line 2:  Add 'of' after 'exchange'.

Yes

> 
> PDF page 9, lines 4-5:  Missing tables.

TBD

> 
> PDF page 9, line 17: 'The _Float16 type need not be an interchange
> floating type.' should(?) be 'The _Float16 type need not be an
> arithmetic floating type.'

One of my biggest worries with Part 3 is that we're affecting the type system. I've used this hierarchy:

floating types
	real floating types
		generic floating types
		decimal floating types (if Part 2 is supported)
		interchange floating types
		extended floating types
	complex types

real types
	integer types
	real floating types

arithmetic types
	integer types
	floating types

interchange types (which include interchange floating types)

Part 3 adds: interchange types, interchange floating types, and extended floating types. I'm trying to use these designations consistently in the draft, and believe the wording about _Float16 is consistent.


> 
> PDF page 9, line 23:  Missing table.

TBD

> 
> PDF page 9, lines 30-31 say real floating is union of generic,
> interchange, and extended.  What about decimal types?

Decimal types are a feature of Part 2 which is not required for Part 3. So someone supporting Parts 2 and 3 (or 1 and 3 or all of them) was to merge the specs. Not ideal, but what's a better alternative?

> 
> PDF page 10, line 5: 'decimal interchange types of with N.' should(?)
> be 'decimal interchange types of width N.'.

Yes

> 
> PDF page 11, line 28:  Need to add xxx_TRUE_MIN macros.

Agreed

> 
> PDF page 12, line 4: Do we need to add words about extended types and
> conversions?

Since extended floating types are real floating types (see hierarchy above) they should already be covered.

> 
> PDF page 12, line 36 has 'real floating type or interchange type'.  I
> think 'interchange type' is not needed.  PDF page 9, lines 30-31 say
> real floating is union of generic, interchange, and extended.

Real floating types include  interchange floating types, but not interchange types.

> 
> PDF page 13, line7:  What is a 'not interchange floating type'?

Any interchange type for which the implementation does provide all the floating-point operations.

> 
> PDF page 13, line 29:  'fNN' -> 'fNx'.

Yes

> 
> PDF page 13, lines 28-30:  Should the 'N' be italic?

Yes

> 
> PDF page 17, lines 24-27 and 45-48:  cbrt is done twice.

Second set should be sqrt.

> 
> PDF page 19, line 28: Should there be _FloatN and _FloatNx remquo
> functions?

Yes (unless we want to deprecate remquo, which seems like a time consuming tangent)

> 
> PDF page 20, lines 2-4: The nandNN functions look wrong.

Yes (incomplete edit)

> 
> 
> ---
> Fred J. Tydeman        Tydeman Consulting
> tydeman at tybor.com      Testing, numerics, programming
> +1 (775) 287-5904      Vice-chair of PL22.11 (ANSI "C")
> Sample C99+FPCE tests: http://www.tybor.com
> Savers sleep well, investors eat well, spenders work forever.
> 
> _______________________________________________
> Cfp-interest mailing list
> Cfp-interest at oakapple.net
> http://mailman.oakapple.net/mailman/listinfo/cfp-interest