Re: floating point precision discussion

Chris Dodd <chrisd@reservoir.com>
22 Sep 1998 22:30:08 -0400

          From comp.compilers
Related articles
Re: inlining + optimization = nuisance bugs luddy@concmp.com (Luddy Harrison) (1998-09-22)
Re: floating point precision discussion jacob@jacob.remcomp.fr (1998-09-22)
Re: floating point precision discussion chrisd@reservoir.com (Chris Dodd) (1998-09-22)
| List of all articles for this month |
From: Chris Dodd <chrisd@reservoir.com>
Newsgroups: comp.compilers
Date: 22 Sep 1998 22:30:08 -0400
Organization: Reservoir Labs
References: 98-09-093 98-09-112
Keywords: arithmetic

Our esteemed moderator wrote:
> [Well, actually, there's are flag bits that lets you set the internal
> precision of the x86 to 32, 64, or 80 bits, so the forced store
> argument is a red herring.


Actually, those flag bits only set the mantissa precision to 24, 53,
or 64 bits. They don't affect the exponent size, so even if you
restrict an x86 to double precision instead of extended, it can still
`fail' by not under/overflowing when it should. This makes these bits
essentially useless if you want strict IEEE semantics. The only way
to get that is to actually store each intermediate to memory and
reload it...


Chris Dodd
chrisd@reservoir.com
--

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.