[Cfp-interest] Not trigger sNaN

[Mike Cowlishaw]

Makes sense .. thanks.
 
Mike


  _____  

From: Eric Schwarz [mailto:eschwarz at us.ibm.com] 
Sent: 17 September 2018 13:26
To: Mike Cowlishaw
Cc: cfp-interest at oakapple.net; 'Jim Thomas'
Subject: RE: [Cfp-interest] Not trigger sNaN



Mike,

I don't remember defining the sameQuantum operation. 
I think it was to compare exponents or as a compare units instruction 
and used it to determine whether the result had the expected quantum. 
This was before IBM architecture came up with a flag for quantum exceptions
and
before we figured out we could test the most significant digit for non-zero.
So if we are just testing a result for correct units, we don't want to
trigger exceptions,
though we may want to force it into the correct units.
An example is we have a table of data to output that should have the units
of 10^-2, we
test each data item before converting the significand to BCD and outputting
the
data. Probably figured SNaNs would have already been seen by this point and
not a point in the program where we want to take any exceptions.

Regards,
Eric


eschwarz at us.ibm.com

"Mike Cowlishaw" <mfc at speleotrove.com> wrote on 09/17/2018 06:47:19 AM:

> From: "Mike Cowlishaw" <mfc at speleotrove.com>
> To: "'Jim Thomas'" <jaswthomas at sbcglobal.net>
> Cc: <cfp-interest at oakapple.net>, "'Eric Schwarz'" <eschwarz at us.ibm.com>
> Date: 09/17/2018 06:47 AM
> Subject: RE: [Cfp-interest] Not trigger sNaN
> 
> I believe it was, but I do not recall the reasoning.  Eric Schwartz might
> remember; cc:ing him ...
> 
> Mike 
> 
> > -----Original Message-----
> > From: Jim Thomas [mailto:jaswthomas at sbcglobal.net] 
> > Sent: 16 September 2018 16:40
> > To: Mike Cowlishaw
> > Cc: cfp-interest at oakapple.net
> > Subject: Re: [Cfp-interest] Not trigger sNaN
> > 
> > Mike,
> > 
> > 754 seems to say sameQuantum(sNaN, qNaN) is true with no 
> > exception. Just double checking that it is as intended.
> > 
> > Jim
> > 
> > > On Aug 28, 2018, at 7:16 PM, Fred J. Tydeman 
> > <tydeman at tybor.com> wrote:
> > > 
> > > On Tue, 28 Aug 2018 13:13:45 -0500 Rajan Bhakta wrote:
> > >> 
> > >>         *AI*: Fred: See which other functions have the need to not 
> > >> trigger signaling nans but are functions (need to be macros).
> > > 
> > > 754 Operations that shall not trigger an sNaN (signal invalid):
> > > 
> > > "*" are functions in C
> > > 
> > > 
> > ...
> > > sameQuantum() = samequantum()*
> > > 
> > 
> 



-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20180917/4e174471/attachment-0001.html