Re: Alias computation? (Preston Briggs)
Mon, 11 Apr 1994 16:05:50 GMT

          From comp.compilers

Related articles
Alias computation? (1994-04-08)
Re: Alias computation? (1994-04-11)
| List of all articles for this month |

Newsgroups: comp.compilers
From: (Preston Briggs)
Keywords: optimize, analysis
Organization: Rice University, Houston
References: 94-04-050
Date: Mon, 11 Apr 1994 16:05:50 GMT writes:
>I need to compute the alias relations for general nested procedures.
>I chose the method of Cooper and Kennedy from: Fast interprocedural
>alias analysis, Sixteenth ACM Symposium on Principles Of Programming
>Languages, 1989. However, I think the adaptation to general nesting
>is wrong, the method is ignoring aliases between two formals of
>different procedures under some circumstances.

It's possible the paper is incorrect; however, it's also possible you're
misunderstanding some point. If you want to pursue it, find a
counter-example and send it to me (along with some commentary explaining
what you think has gone wrong) and I'll get Cooper to look at it and get
back to you.


In general, when I think I've found a problem in a paper, I proceed
through the following sequence until satisfied:

0) Ignore it (maybe, if I don't need the result).

1) Reread the paper, assuming I'm being dumb. Repeat until

2) Ask my local friends about it, generally working up from grad
      students to post-docs to professors.

3) Write the authors. Of course, some authors are quite
      difficult to reach (e.g., Cooper, who never reads e-mail :-)

4) Write a paper.

Posting on comp.compilers might replace step 4 if I was more concerned
with results than papers! (i.e., if I were in industry rather than

Obviously, this is a very traditional approach. The idea is to preserve
everyone's reputations as far as possible. As a side benefit, it cuts
down on the amount of noise.

Preston Briggs

Post a followup to this message

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