Proposal for SentinelInputRange

[Walter Bright]

On 2/28/2013 9:25 AM, Steven Schveighoffer wrote:
> On Thu, 28 Feb 2013 12:00:48 -0500, Walter Bright <newshound2 at digitalmars.com>
> wrote:
>
>> On 2/28/2013 6:31 AM, Steven Schveighoffer wrote:
>>> If this doesn't translate to the same code, I don't know why not.
>>
>> Try it and see with your favorite C compiler.
>
> A sample case of 1 does not prove it's not possible, or explain why those
> optimizers don't take that step.  A valid response would be to give a case why
> an optimizer COULDN'T make that leap.

No, it is not. DMD is compiled with real compilers, not abstract "sufficiently 
smart compilers".

>> Then try the lookahead cases I also posted.
>
> You have already stated it gets changed into a jump table.

Please, please listen to what I write. This is very frustrating. The code in 
lexer.c is there for all to see, and it amply illustrates everything I'm saying. 
For example, this code does not get translated into a jump table:

             case '+':
                 p++;
                 if (*p == '=')
                 {   p++;
                     t->value = TOKaddass;
                 }
                 else if (*p == '+')
                 {   p++;
                     t->value = TOKplusplus;
                 }
                 else
                     t->value = TOKadd;
                 return;

> Such an optimization
> seems possible to me, even with the != 0 check outside the switch, even if not
> all C compilers employ it.

It doesn't matter if it is theoretically possible if the compilers we need to 
use do not do it.