Re: [RFC] A new data structure for SWITCH_EXPR

[Zack Weinberg <zack at codesourcery dot com>]

Andrew Pinski <pinskia@physics.uc.edu> writes:

> On Nov 8, 2004, at 12:15 AM, Zack Weinberg wrote:
>> First, I'd like to point out that use of the "case A ... B:" feature
>> is extremely rare, at least when it's via the GNU extension to C.
>> Perhaps other languages use it more frequently.  As such, it might
>> make sense to drop CASE_HIGH entirely, having front ends break up A
>> ... B into individual CASE_LABEL_NODEs for all the values in the
>> range.  Ranges would get more expensive, but each individual case
>> would get cheaper.
>
> Or are they.
> For an example, look at the following:
>  void f();
>  void g();
>
>  void a(int b) {
>   switch (b) {
>          case 1 ... 2000:
>            f();
>
>          case 2001 ... 4000:
>            g();
>      }
>  }
>
> We would have 4000 case statements in this case which is very
>inefficient as you would then have to look through all 4000 case
>statements instead of just three (one for the default). Yes this is a
>made up case but we do have cases where we have "case '0' ... '9':"
>all the time (yes spelt out but we combine them to save space and do
>expand tricks on them).  (For the '0' ... '9' case we would have to
>do 8 less compares which can help a lot as shown above in the made up
>case).

I think both your examples are artificial.  I think the common case
(at least the common case where the switch is large enough to be a
bottleneck) is lots and lots of DISJOINT case labels.  Think Yacc
parsers, or the bytecode interpreter in libjava.

zw