Re: Prediction of local code modifications

Hi,


>> How do compiler developers cope with this problem? Are there any
>> approaches which allow to predict the influence of a local code
>> optimization on the global code performance for complex processors?
>
> If I understand your question right, the answer is 'dynamic
> programming'.


Thank you for your answer.


I'm not sure if dynamic programming is an approach that I can apply to
my problem. When I understand the idea of dynamic programming
correctly, it exploits the idea of "overlapping subproblems" and
"memoization", i.e. it is assumed that the problem can be divided into
independent subproblems which can be solved separately and then their
optimal solution can be used to construct the global optimal solution.


For my problem with the alignment I could divide the code into smaller
subproblems where I could try to find an optimal local
solution. However, these subproblems are not independent. When I move
some code locally in one place (let's say that's the region of the
first subproblem), then this might possibly also influence some
following code in another region that I consider as a further
subproblem. Thus, calculating separate optimal local solutions and
them combine them will not work for me.


Do you see another way to cope with my optimization problem?


Regards,
Tim

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