Re: Motion P1788/M0029.01: Level-3-interface-only --- Final version to vote on
Baker, and all
On 14 Dec 2011, at 23:37, Dan Zuras Intervals wrote:
> If our compiler people do not object, I think this is an
> excellent idea. The use of "type" as an abstraction &
> "format" as the layout in memory is more in keeping with
> the colloquial use of these words. Therefore, more clear
> to our users.
>
> The only caveat that remains is that you should note when
> you define them that these words are being used differently
> than they were in 754.
Will do.
> It is better to correct tortured prose than propagate it.
The wording of this motion, and use of "type" vs "format" in particular, seem to have the thumbs-up from a number of experts, so I suggest we start the voting period.
John
----------
P.S. Dan's words '... "format" as the layout in memory' aren't quite correct.
... "format" as what the Level 3 programmer sees
is more correct.
E.g., the programmer, writing operations for an IEEE binary64 inf-sup type, knows it stores [xlo,xhi] as xlo and xhi which are binary64 values. They needn't be in that order in memory, or even contiguous. Depending on the language and the design of the library, she might access them, for an interval object xx, as
xx.xlo and xx.xhi
or
xx%INF and xx%SUP
etc. In C/C++, I believe they could actually be stored, say, as (-xlo) and xhi, with a preprocessor "def" that makes this invisible.