Re: recognition and optimization of prefix computation or variants in C/C++ compilers

rcmetzger <rcmetzger@grandecom.net>
Mon, 23 Jun 2008 07:36:20 -0700 (PDT)

          From comp.compilers
Related articles
recognition and optimization of prefix computation or variants in C/C+ sandyasm@gmail.com (2008-06-17)
Re: recognition and optimization of prefix computation or variants in gneuner2@comcast.net (George Neuner) (2008-06-18)
Re: recognition and optimization of prefix computation or variants in idbaxter@semdesigns.com (2008-06-21)
Re: recognition and optimization of prefix computation or variants in rcmetzger@grandecom.net (rcmetzger) (2008-06-23)
| List of all articles for this month |
From: rcmetzger <rcmetzger@grandecom.net>
Newsgroups: comp.compilers
Date: Mon, 23 Jun 2008 07:36:20 -0700 (PDT)
Organization: Compilers Central
References: 08-06-035
Keywords: optimize
Posted-Date: 23 Jun 2008 10:36:46 EDT

On Jun 17, 3:34 am, sandy...@gmail.com wrote:
> Does the current C/C++ compilers in industry recognize idioms of the
> form of prefix computation and transform them? For instance, given
> If so, under what class of optimizations do they do this?
>
> In the general case, a similar prefix pattern can be identified and
> seen in certain search/traversal algorithms also. Do the existing
> compilers handle any of those as well?


I have worked on the phase of an optimizer for Fortran/C/C++ that
recognized some prefix computations with ad hoc pattern matching.


See the Introduction to "Automatic Algorithm Recognition and
Replacement: A New Approach to Program Optimization" (Metzger and Wen,
MIT Press, 2000) for an explanation of the limitations of ad hoc
pattern matching in optimizing compilers.


Read the rest of the book for an alternative that generally doesn't
have those limitations.


Robert Metzger
Sr. Compiler Engineer
Convey Computer Corp.

Post a followup to this message

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