Re: Recognize valid paths

Tim Frink <>
26 Aug 2008 15:30:34 GMT

          From comp.compilers

Related articles
Recognize valid paths (Tim Frink) (2008-08-20)
Re: Recognize valid paths (2008-08-23)
Re: Recognize valid paths (Hans-Peter Diettrich) (2008-08-24)
Re: Recognize valid paths (Tim Frink) (2008-08-26)
Re: Recognize valid paths (Tim Frink) (2008-08-26)
Re: Recognize valid paths (Michiel Helvensteijn) (2008-08-27)
Re: Recognize valid paths (Jeff Kenton) (2008-09-01)
| List of all articles for this month |

From: Tim Frink <>
Newsgroups: comp.compilers
Date: 26 Aug 2008 15:30:34 GMT
Organization: Compilers Central
References: 08-08-042 08-08-053
Keywords: analysis, optimize
Posted-Date: 26 Aug 2008 23:34:13 EDT

> Intuitively I'd partition the values of x into <1, 1..10 and >10, based
> on the tested conditions. Then subdivide these ranges, when required by
> modifications of the value in some branch (this will not happen in your
> example). Finally I'd check whether for every partition a single path is
> taken, what again is true in your example. If not, the analysis may
> deserve more table space for the possible pathes.

Will this be not complex for real-world applications that might have
a huge number of paths?

> Another inspection of
> the ever visited BB's, in all pathes, will reveal dead code - if this is
> what you want to know.

This is an information about the code that is independent of the calling
context since these BB's are never reached during program execution.

> Otherwise I don't understand what an "invalid"
> path (never taken) should be useful to know. Branch and bound
> (wikipedia)?

The problem to know which paths are taking in which context are
for example crucial for time estimations based on a static program


Post a followup to this message

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