This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: CSE optimization question



On Thursday, January 10, 2002, at 09:36  AM, law@redhat.com wrote:

>> I asked this before, but why isn't GCSE making that obsolete now? It 
>> isn't,
>> but I don't understand *why* it isn't.
> There's a lot of code in cse that doesn't exist in gcse -- for example 
> cse
> does a lot of simplifications that gcse knows nothing about.
>
> But more importantly gcse is based on common subexpression elimination
> while cse is actually a derivative of value numbering.
Right.
In fact, it does little CSE, if any, at all.
>  Meaning they're
> based on completely different concepts for identifying expressions which
> are redundant.  And finally, gcse doesn't handle redundancies which 
> occur
> within a single basic block.

This could be fixed pretty easily by doing it during local availability 
computation, no?
>
> jeff


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]