Re+ :Perl Idioms Explained: && and || "Short Circuit" operators

[BrowserUk]

Modern C compilers have the smarts built in to make intelligent decisions on what code to emit. In the case of your example of 2 widely spaced cases, it produces this

?live1@0:
   ;    
   ;    int main( int argc, char * argv[] ) {
   ;    
    push      ebp
    mov       ebp,esp
    push      ecx
   ;    
   ;        unsigned long selector;
   ;        sscanf( argv[1], "%li", &selector );
   ;    
@1:
    lea       eax,dword ptr [ebp-4]
    push      eax
    push      offset s@
    mov       edx,dword ptr [ebp+12]
    push      dword ptr [edx+4]
    call      _sscanf
    add       esp,12
   ;    
   ;    
   ;        switch ( selector ) {
   ;    
    mov       ecx,dword ptr [ebp-4]
    sub       ecx,-2147483648
    je        short @5
    sub       ecx,-2147483647
    jne       short @2
   ;    
   ;            case 1            : printf( "case 1\n" );             
+break;
   ;    
    push      offset s@+4
    call      _printf
    pop       ecx
@8:
    pop       ecx
    pop       ebp
    ret 
   ;    
   ;            case 2147483648    : printf( "case 2147483648\n" );   
+ break;
   ;    
@5:
    push      offset s@+12
    call      _printf
    pop       ecx
@9:
    pop       ecx
    pop       ebp
    ret 
   ;    
   ;            default            : printf( "default\n" );           
+ break;
   ;    
@2:
    push      offset s@+29
    call      _printf
    pop       ecx
   ;    
   ;        }
   ;    }
   ;    
@6:
@7:
    pop       ecx
    pop       ebp
    ret 
_main    endp
_TEXT    ends

Which as you can see, gets translated into a test and jump for the extreme case, another test and jump for the default case and a fall through for the 1 case.

If you combine a contiguous range and an extreme case, you get.

_TEXT    segment dword public use32 'CODE'
_main    proc    near
?live1@0:
   ;    
   ;    int main( int argc, char * argv[] ) {
   ;    
    push      ebp
    mov       ebp,esp
    push      ecx
   ;    
   ;        unsigned long selector;
   ;        sscanf( argv[1], "%li", &selector );
   ;    
@1:
    lea       eax,dword ptr [ebp-4]
    push      eax
    push      offset s@
    mov       edx,dword ptr [ebp+12]
    push      dword ptr [edx+4]
    call      _sscanf
    add       esp,12
   ;    
   ;    
   ;        switch ( selector ) {
   ;    
    mov       ecx,dword ptr [ebp-4]
    cmp       ecx,3
    jg        short @11
    je        short @7
    sub       ecx,-2147483648
    je        short @10
    sub       ecx,-2147483647
    je        short @9
    dec       ecx
    je        short @8
    jmp       short @2
@11:
    sub       ecx,4
    je        short @6
    dec       ecx
    je        short @5
    dec       ecx
    je        short @4
    jmp       short @2
   ;    
   ;            case 1            : printf( "case 1\n" );             
+break;
   ;    
@9:
    push      offset s@+4
    call      _printf
    pop       ecx
@14:
    pop       ecx
    pop       ebp
    ret 
   ;    
   ;            case 2            : printf( "case 2\n" );             
+   break;
   ;    
@8:
    push      offset s@+12
    call      _printf
    pop       ecx
@15:
    pop       ecx
    pop       ebp
    ret 
   ;    
   ;            case 3            : printf( "case 3\n" );             
+   break;
   ;    
@7:
    push      offset s@+20
    call      _printf
    pop       ecx
@16:
    pop       ecx
    pop       ebp
    ret 
   ;    
   ;            case 4            : printf( "case 4\n" );             
+   break;
   ;    
@6:
    push      offset s@+28
    call      _printf
    pop       ecx
@17:
    pop       ecx
    pop       ebp
    ret 
   ;    
   ;            case 5            : printf( "case 5\n" );             
+   break;
   ;    
@5:
    push      offset s@+36
    call      _printf
    pop       ecx
@18:
    pop       ecx
    pop       ebp
    ret 
   ;    
   ;            case 6            : printf( "case 6\n" );             
+   break;
   ;    
@4:
    push      offset s@+44
    call      _printf
    pop       ecx
@19:
    pop       ecx
    pop       ebp
    ret 
   ;    
   ;            case 2147483648    : printf( "case 2147483648\n" );   
+ break;
   ;    
@10:
    push      offset s@+52
    call      _printf
    pop       ecx
@20:
    pop       ecx
    pop       ebp
    ret 
   ;    
   ;            default            : printf( "default\n" );           
+ break;
   ;    
@2:
    push      offset s@+69
    call      _printf
    pop       ecx
   ;    
   ;        }
   ;    }
   ;    
@12:
@13:
    pop       ecx
    pop       ebp
    ret 
_main    endp
_TEXT    ends

Which has been converted into a test cascade something like the original Perl code. However, this is before the optimiser has gone to work on it. Once the optimiser sees that code, it is quite likely that it will revert to using a dispatch table for the continuous cases with a single test for the extreme case. I having having trouble obtaining a post - optimisation disassembly that is readable. I'll try to update with that later.

As you might guess, if your range of cases is a widely spread set of discontiguous cases, then the result is you end up with cascaded tests just like the perl version. However, in almost every case (sic:) I remember using, I found some fairly simple function theat mapped the range to a contiguous or near-contiguous integer range. When this isn't the case, then I use an if/else cascade instead.

---

Examine what is said, not who speaks.

"Efficiency is intelligent laziness." -David Dunham
"Think for yourself!" - Abigail
Hooray!

[demerphq]

So the rub of this is that switch() is not O(1). But that as often as possible the optimizer will do its best to convert it to be so. Which if you think about it means that from a programmers perspective its merely syntactic sugar for a chain of ifs. I say this becuase its just as reasonable to assume that the optimizer could figure out that a chain of ifs can be represented by a jump table and do the same optimization there.

Also in your private message to me you mentioned that on occassion the compiler constructs a multi gigabyte executable to try to provide this "optimization". Which I would have thought _totally_ defeats the purpose given that the swap/thrash times involved to implement it would _far_ outweigh the execution times of a chained if.

Anyway, many thanks to you and the other respondants about this. I have certainly learned something.

[TimToady]

I say this becuase its just as reasonable to assume that the optimizer could figure out that a chain of ifs can be represented by a jump table and do the same optimization there.

Indeed, Perl4 did precisely that. We kept threatening to do it in Perl5, but nobody ever got around to doing it. Perl6 will need to do it--the switch statement was designed with the assumption that it would be optimized in the simple cases.