This is the mail archive of the gcc-patches@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: [4.5] Doloop improvement patches, 2/7


Hi,

> Zdenek Dvorak wrote:
> 
> > hmm... that certainly is rather counterintuitive.  I'd recommend to just
> > drop did_replace, the performance gain is probably negligible compared
> > to the maintentance nightmare potential :-)
> 
> Actually I'd prefer to keep it, as doing unnecessary work in this code
> could also lead to confusion.  I'm open to better naming suggestions.

in that case, please split simplify_using_assignment to two functions:
one that checks whether the assignment has the correct form, and the other
one that performs the replacement.  I.e., the code would become
something like 

if (usable_assignment_p (insn))
  {
    simplify_using_assignment (insn, expr);
    for (cond in conds)
      simplify_using_assignment (insn, cond)
  }

> > What about removing the conditions invalidated by this_altered?
> > I guess it is not actually needed now because simplify_using_condition
> > tests the set of altered registers, but it might be nicer to
> > deal with the registers only in simplify_using_initial_values (similarly
> > like you did for simplify_using_assignment),
> 
> This is something I'd like to clean up as well, but it seems more or
> less unrelated.  Something for 8/7 maybe.

Sure, that's ok with me,

Zdenek


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