From: linuxkaffee_@_gmx.net   
      
   > I would say the contrary: scheduling based on profiling information has   
   > no practical sense, since profiling information depends on the chosen   
   > data input. For numerical applications, this would not be really a   
   > problem, and profiling helps. But for general purpose applications,   
   > profiling does not necessarily helps to produce better codes for any   
   > data input.  This is why static code optimisation does not rely on   
   > profiling.   
      
   OK, so taking the danger of changing hot paths through different inputs   
   into account, do you think that it might be a promising idea to mix   
   instructions from several paths? This would possibly help the scheduler   
   to reorder instructions such that the pipelines/functional units are   
   maximally exploited but on the other hand no particular path is favored/   
   neglected too heavily.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)