Re: usqrtsafe isn't

Bill Allombert on Tue, 17 Nov 2009 18:12:36 +0100

[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: usqrtsafe isn't

To: pari-dev@list.cr.yp.to
Subject: Re: usqrtsafe isn't
From: Bill Allombert <Bill.Allombert@math.u-bordeaux1.fr>
Date: Tue, 17 Nov 2009 16:26:36 +0100
Delivery-date: Tue, 17 Nov 2009 18:12:36 +0100
In-reply-to: <20091116185027.GD4458@yellowpig>
Mail-followup-to: pari-dev@list.cr.yp.to
Mailing-list: contact pari-dev-help@list.cr.yp.to; run by ezmlm
References: <11af68b20911152128l4ed06651x7e5de1dc8320ae59@mail.gmail.com> <20091116185027.GD4458@yellowpig>
User-agent: Mutt/1.5.18 (2008-05-17)

On Mon, Nov 16, 2009 at 07:50:27PM +0100, Bill Allombert wrote:
> On Mon, Nov 16, 2009 at 12:28:19AM -0500, Charles Greathouse wrote:
> > In writing a program using the Pari library today I was able to trace
> > a problem to usqrtsafe.  In particular,
> > usqrtsafe(18446744073709551615) = 4294967296
> > but 4294967296^2 = 18446744073709551616 > 18446744073709551615.
> 
> Thanks a lot for reporting this problem!
> 
> The code suppose the error is always by default by actually it is always
> by excess, because the conversion to 'double' round to the nearest and
> usqrtsafe should round to the lowest.
> 
> The patch below should fix that.

I commited this patch as SVN revision 12026.

Cheers,
Bill.

References:
- usqrtsafe isn't
  - From: Charles Greathouse <charles.greathouse@case.edu>
- Re: usqrtsafe isn't
  - From: Bill Allombert <Bill.Allombert@math.u-bordeaux1.fr>

Prev by Date: Re: usqrtsafe isn't
Previous by thread: Re: usqrtsafe isn't
Index(es):
- Date
- Thread