Re: Re: genetic compilation

"felix" <felix@anu.ie>
11 Jun 2000 12:31:30 -0400

          From comp.compilers
Related articles
Re: Re: genetic compilation rbw3@dana.ucc.nau.edu (Brock) (2000-04-01)
Re: Re: genetic compilation joachim.durchholz@halstenbach.com.or.de (Joachim Durchholz) (2000-04-14)
Re: Re: genetic compilation felix@anu.ie (felix) (2000-06-11)
Re: Re: genetic compilation joachim.durchholz@halstenbach.com.or.de (Joachim Durchholz) (2000-06-14)
| List of all articles for this month |
From: "felix" <felix@anu.ie>
Newsgroups: comp.compilers
Date: 11 Jun 2000 12:31:30 -0400
Organization: Posted via Supernews, http://www.supernews.com
References: 00-04-019 00-04-109
Keywords: optimize

>> > I once read something about compilers that tried out equivalent
>> > machine code sequences and measured actual performance to decide
>> > which worked best. This could be used as a basis for a genetic
>> > algorithm. Of course, [...] it wouldn't work for cross-compilation
>> > anyway.


(Sorry, I haven't followed this thread so my question may be redundant)


I thought for some time about compilation with support of genetic
algorithms and I would like to know wether there is currently any
ongoing research. Has someone tried this out, yet? Generating native
code and mucking around with it in a GA has to be done extremely
careful, to avoid breaking constraints on data-representation or
execution-order.


To summarize: How can I prove that a code-sequence that is genetically
manipulated handles all legal inputs properly that are given to it?


felix

Post a followup to this message

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