Re: Determining instruction mix

mueller@nu.cs.fsu.edu (Frank Mueller)
Wed, 8 Feb 1995 17:43:42 GMT

          From comp.compilers

Related articles
Determining instruction mix s9154221@mella.ee.up.ac.za (1995-02-06)
Re: Determining instruction mix nandu@cs.clemson.edu (1995-02-07)
Re: Determining instruction mix gardner@pink-panther.cs.uiuc.edu (1995-02-07)
Re: Determining instruction mix mueller@nu.cs.fsu.edu (1995-02-08)
Re: Determining instruction mix pardo@cs.washington.edu (1995-02-09)
Re: Determining instruction mix roos@grendel.csc.smith.edu (1995-02-09)
| List of all articles for this month |

Newsgroups: comp.compilers
From: mueller@nu.cs.fsu.edu (Frank Mueller)
Keywords: architecture, testing
Organization: Florida State University Computer Science
References: 95-02-068 95-02-081
Date: Wed, 8 Feb 1995 17:43:42 GMT

Rikus Combrinck (s9154221@mella.ee.up.ac.za) wrote:
: I'm looking for software that can determine the dynamic
: instructions mix of a programme, ...
: [Somewhere I was a paper about a system that does this by instrumenting
: basic blocks so it makes one count per block rather than per instruction,
: then post-proccesses to get the instruction counts. -John]


>, gardner@pink-panther.cs.uiuc.edu (Mark K. Gardner) writes:
|> I seem to remember that the folks at University of Wisconsin - Madison
|> had a system to instrument basic blocks (as John pointed out). Their
|> purpose was to generate memory reference traces, so it may not do exactly
|> what you had in mind. But then again, maybe it could be extended.


Check out the following references:


@inproceedings{ball,
                author = "T. Ball and J. R. Larus",
                title = "Optimally Profiling and Tracing Programs",
                booktitle = POPL,
                year = "1992",
                month = jan,
                pages = "59-70"
                }
also published in extended format and detail in ACM TOPLAS last year. Shows that
a subset of basic blocks is sufficient for profiling / tracing. They also have a
tool describe in the following reference, I believe:
@article{larus93,
                author = "J. R. Larus",
                title = "Efficient Program Tracing",
                journal = "IEEE Computer",
                volume = 26,
                number = 5,
                year = "1993",
                month = may,
                pages = "52-61"
                }




@inproceedings{davidson90,
                author = "J. W. Davidson and D. B. Whalley",
                title = "Ease: An Environment for Architecture Study and
                                  Experimentation",
                booktitle = SIGMETRICS,
                year = "1990",
                month = may,
                pages = "259-260"
                }
Describes how to integrate basic block profiling with a compiler back-end.


@inproceedings{mueller94-3,
                author = "F. Mueller and D. B. Whalley",
                title = "Efficient On-the-fly Analysis of Program Behavior
                                  and Static Cache Simulation",
                booktitle = "Static Analysis Symposium",
                series = LNCS,
                volume = 864,
                publisher = "Springer",
                editor = "B. Le Charlier",
                year = "1994",
                month = sep,
                pages = "101-115"
                }
Describes how instrumentation can be reduced for order-dependent on-the-fly-
analysis, e.g. for caching. (ftp via my WWW page for publications, see below.)


Hope this help,


--
Frank Mueller E-Mail: mueller@cs.fsu.edu WWW: http://www.cs.fsu.edu/~mueller
--


Post a followup to this message

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