To be fair, the article does specifically say this in the context when "case labels are placed far apart".
This is would lead to a "sparse" lookup table, where most of the entries would be unused; eg,

switch(x)
{
   case   1: // do stuff
             break;
   case 100: // do stuff
             break;
   case 200: // do stuff
             break;
   case 255: // do stuff
             break;
}

A decent modern optimising compiler should probably be smart enough to spot this, and effectively break the the statement up itself, or use an if-else-if cascade.


True but, in the example above, the code would be very fast but very big!

This just bings us back to the same old story: there is no such thing as "Optimum" or "Best" - it always has to be qualified; eg, "optimum" for speed will often mean sub-optimum for code size; "best" for code size will often mean sub-optimum for speed! :-(
Such is life!

Remember Red Adair: http://www.8052.com/forum/read.phtml?id=90840