If statements vs case

[Mark Smith]

Richard, I bow to your more extensive test. All I did was a simple  
five-way switch/if (based on a random input)  that actually did  
nothing, so I think your test is probably more useful.

Best,

Mark

On 26 Feb 2007, at 17:35, Richard Gaskin wrote:

> Mark Smith wrote:
>> I just ran a very simple benchmark test, which suggests that if/ 
>> then  goes about 20% faster than switch/case.
>
> I get different results here, but I had to rewrite the if-then test  
> as the example posted as it wasn't executing the same logic as the  
> switch block.  In the "if" example the "c" case didn't fall through  
> to also execute the "d" case as it did in the "switch" example, and  
> the "e" case wasn't handled at all.
>
> So I wrote a fresh pair of examples that test all possible cases,  
> running each full set 10000 iterations. They each produce the same  
> logical result at this point, with these times on my MacBook Pro:
>
> if:89ms switch:74ms
> if result:140000 switch result:140000
>
> The timings aren't surprising given the extra evaluations needed to  
> get the same effect using "if" as with the "switch" fall-through  
> behavior.
>
> Please double-check the benchmark handler below to make sure that  
> both tests are using equivalent logic, and look for ways the "if- 
> then" might be better optimized:
>
>
>
> local sResult
>
> on mouseUp
>   -- number of test iterations:
>   put 10000 into n
>   -- load var with all possible values:
>   put "abcde" into s
>   --
>   --
>   -- TEST 1: if-then
>   put 0 into sResult
>   put the millisecs into t
>   repeat n
>     --
>     repeat for each char tVar in s
>     If tVar ="a" or tVar ="b" then
>         DoThing1
>       else if tVar = "c" or tVar = "d" then
>         if tVar = "c" then
>           DoThing2
>         end if
>         DoThing3
>       Else If tVar ="e" then
>         DoThing4
>       End if
>     end repeat
>     --
>   end repeat
>   put the millisecs - t into t1
>   put sResult into tResult1
>   --
>   -- TEST 2: switch
>   put 0 into sResult
>   put the millisecs into t
>   repeat n
>     --
>     repeat for each char tVar in s
>       switch tVar
>       case "a"
>       case "b"
>         DoThing1
>         break
>       case "c"
>         DoThing2
>       case "d"
>         DoThing3
>         break
>       case "e"
>         DoThing4
>       end switch
>     end repeat
>     --
>   end repeat
>   put the millisecs - t into t2
>   put sResult into tResult2
>   --
>   -- show results:
>   put "if:"&t1 &"ms"&&"switch:"& t2&"ms"&cr&\
>  "if result:"& tResult1 &&"switch result:"&tResult2
> end mouseUp
>
>
>
> on DoThing1
>   add 1 to sResult
> end DoThing1
>
> on DoThing2
>   add 2 to sResult
> end DoThing2
>
> on Dothing3
>   add 3 to sResult
> end Dothing3
>
> on DoThing4
>   add 4 to sResult
> end DoThing4
>
>
>
>
> -- 
>  Richard Gaskin
>  Managing Editor, revJournal
>  _______________________________________________________
>  Rev tips, tutorials and more: http://www.revJournal.com
> _______________________________________________
> use-revolution mailing list
> use-revolution at lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your  
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-revolution