Re: [Texmacs-dev] Efficiency of C++ switch-statements

[Gabriel Dos Reis]

[ I forgot to add the following in my reply ]


"Dan Martens" <address@hidden> writes:

| >Not always.  The compiler (g++ or gcc) can use a table with entries or a
| >series of compares and jumps, that depends on the density of the labels.  
| >See  Message-ID: <address@hidden>
| 
| Yes, this is true, but as Joris stated previously, his switch
| statement was sparse, with many "fall-throughs" to a single piece of
| code. This is what I experimented with.  I apologize if I did not
| cover all the different implementations of a switch statement (of
| which there are many), but I was just answering the question at
| hand. 

   It is not clear to me that Joris' switch is sparse.  I would expect
something  like

   switch (tag) {
     <all different cases>
 
     case similar1:
     case similar2:
     case similar3:
        ....
     case similarN:
      // common code ...
      break;

     default:
       // should not happen
      std::abort();
      break;
   };

which is not what I call a sparse switch-statement even if there are
about 140 fall-throughs.

-- Gaby