Hey Philip,
The switch opcode is constant, and doesn't compare for each target.
But a C# switch can go linear, as:
switch (foo) {
case 42: return 1;
case 74: return 2;
case 96: return 3;
}
will be compiled as:
if (foo == 42) goto case_42;
if (foo == 74) goto case_74;
if (foo == 96) goto case_96;
goto default;
case_42: return 1;
case_74: return 2;
case_96: return 3;
default:
Other than that, you should be limited to Int32.MaxValue for the
number of switch targets.
On 4/17/09, Philip_L <[email protected]> wrote:
>
> Thanks Jb. BTW, do you know if switch statements in IL are done in
> linear or constant time? For example, if I have an a set of
> instruction labels as jump targets in an array using the
> OpCodes.Switch instruction and I already have the target index
> computed at runtime, will it still have to do a comparison for each
> case? And is there a limit on how many labels one can have in a switch
> statement done in IL?
>
>
> On Apr 18, 12:57 am, Jb Evain <[email protected]> wrote:
> > Hey Philip,
> >
>
> > On 4/6/09, Philip_L <[email protected]> wrote:
> >
> > > switch(someString)
> > > {
> > > case "a":
> > > break;
> > > case "b":
> > > break;
> > > default:
> > > break;
> >
> > That's compiled using a Dictionary<>. A switch is quite complex
> > actually. If you have:
> > int a;
> > switch (a) {
> > case 0:
> > case 1:
> > case 12:
> > case 74:
> >
> > }
> >
> > It's likely that only the case 0 and 1 will be compiled using a switch
> > opcode, the rest will be compiled as a short serie of comparison. I
> > suggest you write a few examples and see in ildasm what csc emits,
> > it's quite interesting (I had to do that for Cecil.Decompiler, it's
> > actually not that fun to reconstruct a switch statement from what is
> > actually compiled).
> >
> > --
>
> > Jb Evain <[email protected]>
> >
>
--
Jb Evain <[email protected]>
--~--~---------~--~----~------------~-------~--~----~
--
mono-cecil
-~----------~----~----~----~------~----~------~--~---
