Generated code depends on amount of memory

Derek M Jones <derek@NOSPAMknosof.co.uk>
28 May 2005 14:03:03 -0400

          From comp.compilers

Related articles
Generated code depends on amount of memory derek@NOSPAMknosof.co.uk (Derek M Jones) (2005-05-28)
Re: Generated code depends on amount of memory bstecher@qnx.com (2005-05-31)
Re: Generated code depends on amount of memory bobduff@shell01.TheWorld.com (Robert A Duff) (2005-05-31)
| List of all articles for this month |

From: Derek M Jones <derek@NOSPAMknosof.co.uk>
Newsgroups: comp.compilers
Date: 28 May 2005 14:03:03 -0400
Organization: Knowledge Software
Keywords: code, question
Posted-Date: 28 May 2005 14:03:03 EDT

All,


I have heard about a compiler that takes account of the amount of
memory available in the machine it is being executed on when deciding
how much optimization to do (i.e., more memory gives it more room to
do more optimizations).


The issue of a compiler generating different code from the same source
is an important one in the domain of high integrity applications. I
have heard of compilers that had user selectable options that allowed
developers to tell the compiler how much memory it can use. But I had
not heard of any compilers that would generate different code if the
same command line options were used (I am aware of recent research
that uses genetic algorithms during optimizations).


Does anybody have any knowledge of such compilers?
[Sure. I've run into one where the generated code depended on hash
table values based on memory locations, so if on two runs the
addresses were slightly different as they might be if the environment
were different, you'd get different code. The two versions of the
code were the same size and speed, but they were different. -John]



Post a followup to this message

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