[J3] Question on TINY example

Malcolm Cohen malcolm at nag-j.co.jp
Wed Jun 22 22:15:15 UTC 2022

Hi Vipul,


The description is correct. It follows from the model in 16.4. That model has 1/radix <= mantissa <1. So the minimum number is radix**(emin-1) QED. (This is not how IEEE-754 describe themselves, they have 1<=mantissa<radix.)


To all: these descriptions were written back in Fortran 90 days by numerical experts. That does not mean that they always got it right, but you should think twice, thrice, and then again before assuming that they are wrong.




..............Malcolm Cohen, NAG Oxford/Tokyo.


From: J3 <j3-bounces at mailman.j3-fortran.org> On Behalf Of Vipul Parekh via J3
Sent: Thursday, June 23, 2022 6:41 AM
To: General J3 interest list <j3 at mailman.j3-fortran.org>
Cc: Vipul Parekh <parekhvs at gmail.com>
Subject: Re: [J3] Question on TINY example




On Wed, Jun 22, 2022 at 2:02 PM Steve Lionel via J3 <j3 at mailman.j3-fortran.org <mailto:j3 at mailman.j3-fortran.org> > wrote:

A paper would be appreciated.




Per Steve's note, do you (and the Nvidia team) plan to submit a paper on this?  That will be useful.


More than the example with TINY (and also with HUGE any other involving the "model" in the Note), I am concerned about the description itself.


You will note for TINY, the standard text has for Result Value, "The result has the value b (to the power of ) emin−1 ..".


Should it not be b (to the power of) emin?  The edit to the example can then follow what is consistent with the description of the Result Value.



Vipul Parekh


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.j3-fortran.org/pipermail/j3/attachments/20220623/7f478c30/attachment.htm>

More information about the J3 mailing list