Re: Banerjee inequality

=?ISO-8859-1?Q?Pertti_Kellom=E4ki?= <pertti.kellomaki@tut.fi>
Thu, 08 Nov 2007 12:14:42 +0200

          From comp.compilers
Related articles
Banerjee inequality pertti.kellomaki@tut.fi (=?ISO-8859-1?Q?Pertti_Kellom=E4ki?=) (2007-11-02)
Re: Banerjee inequality rcmetzger@grandecom.net (rcmetzger) (2007-11-04)
Re: Banerjee inequality jle@ural.owlnet.rice.edu (2007-11-05)
Re: Banerjee inequality gneuner2/@/comcast.net (George Neuner) (2007-11-05)
Re: Banerjee inequality pertti.kellomaki@tut.fi (=?ISO-8859-1?Q?Pertti_Kellom=E4ki?=) (2007-11-06)
Re: Banerjee inequality pertti.kellomaki@tut.fi (=?ISO-8859-1?Q?Pertti_Kellom=E4ki?=) (2007-11-08)
| List of all articles for this month |
From: =?ISO-8859-1?Q?Pertti_Kellom=E4ki?= <pertti.kellomaki@tut.fi>
Newsgroups: comp.compilers
Date: Thu, 08 Nov 2007 12:14:42 +0200
Organization: Compilers Central
References: 07-11-006 07-11-016
Keywords: analysis
Posted-Date: 08 Nov 2007 14:48:31 EST

George Neuner wrote
> David Klappholz, Kleanthis Psarris, and Xiangyun Kong, "On the perfect
> accuracy of an approximate subscript analysis test", ACM SIGARCH
> Computer Architecture News , Proceedings of the 4th international
> conference on Supercomputing ICS '90, Volume 18 Issue 3b, June 1990.


Thanks, this is a good source. It also provides an interesting
reason why the Banerjee test works so well in practice: "if there
is a coefficient of +1 or -1, and if all coefficients are smaller,
in absolute value, than all loop iteration ranges, then the Banerjee
test checks for integer solutions".
--
Pertti


Post a followup to this message

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