[Cfp-interest 2051] Re: Subnormals

[Steve (Numerics) Canon]

> On Jul 5, 2021, at 6:48 PM, Jim Thomas <jaswthomas at sbcglobal.net> wrote:
> 
>>> 
>>> "Flush to zero" is not defined and is used only once in C, in a footnote. A related problem is that the operations that flush
>>> results to zero would need to be identified (or characterized). For example, would negate (-), fabs, and copysign be
>>> expected to flush to zero? How about conversion to same or wider format?
>> 
>> I would expect all FP operations to be consistent on flushing; either all or none.
> 
> The ARM documentation says VABS and VNEG signal no exceptions. This suggests that they might not handle subnormals like other floating-point instructions.

Right. FABS and FNEG (the more modern ARM names for these) bind the 5.5.1 operations negate() and abs(), and therefore signal no exceptions and only affect the sign bit (and hence cannot flush subnormal inputs or results).

– Steve
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20210706/da60f7df/attachment.htm>