Talk:Extreme floating point values: Difference between revisions
m
→Humbled
m (→Humbled) |
|||
| (4 intermediate revisions by 2 users not shown) | |||
Line 20:
I have marked the first, Perl 6, as incomplete and will do others as time permits. (Unless what I propose is not wise ). --[[User:Paddy3118|Paddy3118]] ([[User talk:Paddy3118|talk]]) 07:11, 3 October 2015 (UTC)
==Humbled==
I just re-read the Fortran and Ada entries. They are good, aren't they. RC doesn't have a method to flag particularly informative entries such as these; (but if it did, who would police attribution etc...) --[[User:Paddy3118|Paddy3118]] ([[User talk:Paddy3118|talk]]) 09:31, 16 April 2018 (UTC)
:Well... No. As to the Fortran entry, I don't regard verbosity as quality. Older details regarding long-defunct machines are easily found on Bisavers, and are not specific to Fortran (and writing a good summary would require much more work). Negative zero is introduced, but we don't know why (''Certain calculations are said to benefit from the states "positive zero", and "negative zero" being available'', that can't be serious), and I don't know were the authors has found a link with theoretical differentiation. All of this is just hovering over the subject. If you want a good account, find good books or articles about floating-point (Goldberg of course, but also Higham, Golub & Van Loan, Muller...). Particularly informative, this? No, I definitely don't think so. Anyway, I don't think this is the right place for a course on floating point: RC is good but is not the right format for this, or at least not in "task space". [[User:Eoraptor|Eoraptor]] ([[User talk:Eoraptor|talk]]) 18:06, 16 April 2018 (UTC)
::We are of differing opinion; on the web!
::So many languages have grown up with IEEE 754 support as standard. It is good to know a little more about languages that could support differing FP representations (and that might still be asked to do that on current hardware). --[[User:Paddy3118|Paddy3118]] ([[User talk:Paddy3118|talk]]) 19:16, 16 April 2018 (UTC)
:::Most standardized languages assume as little as possible about the floating-point representation (Fortran, C, Ada for instance). Most scripting languages make no specific assumption and take the C mlib for granted. Actually, except Java, I know no language that explicitly states that FP must follow the IEEE 754 standard (and it's not mandatory either in Java). Thus, almost all languages "can support differing FP representation". However, it is very interesting to have an account of historical machines (and not only about FP). But here it's buried in a task that is only slightly related, in a specific language although we are talking about the machines. It would be better to have a good explanation on a separate page, as it's absolutely not specific to one language. Even in the old days, Fortran was never alone - though it was certainly the main language for number crunching. Regarding Fortran entries in general: one contributor enjoys giving many details about defunct machines, and this is sad because, again, it's not about Fortran, and also it gives the impression that Fortran is defunct too. But it's wrong: there have been four ISO standards since Fortran 77, and the fifth is about to be published. Historical details about obscure functionality of antiquated compilers may be interesting, but details about how to write good programs in Fortran 2008 (or Fortran 2018) would be even more interesting, if we assume those who come on RC wants to learn usable material. [[User:Eoraptor|Eoraptor]] ([[User talk:Eoraptor|talk]]) 07:41, 18 April 2018 (UTC)
| |||