little question about how improve a code

09-12-2007, 03:27 PM	#1
Troll-Brain	I've found this : http://www.wc3jass.com/viewtopic.php...7c696f44ae3f43 PitzerMike had said that : // for speed reasons i don't loop through charmaps here. Ok but what about the many if/then/endif ? If i use only one if/then/endif and many elseif, will it be faster ?

09-12-2007, 03:40 PM	#2
Captain Griffen	My guess is that it would compile to exactly the same bytecode.

09-12-2007, 03:44 PM

Troll-Brain

Quote:

Originally Posted by Captain Griffen

My guess is that it would compile to exactly the same bytecode.

Maybe but how can i test it, with the command /fps and a short timer periodic ?

09-12-2007, 04:40 PM	#4
Captain Griffen	There is no way to test the speed difference, but I'm 99% sure that there isn't one.

09-12-2007, 04:45 PM

Troll-Brain

Quote:

Originally Posted by Captain Griffen

There is no way to test the speed difference, but I'm 99% sure that there isn't one.

and if i see the using % of the processor ?

09-12-2007, 06:29 PM	#6
Anitarf	No matter what you do, the difference is probably too small to notice. Maybe with stopwatch natives, but even that would probably be inconclusive due to error margins.

09-12-2007, 06:55 PM	#7
Captain Griffen	The difference, if any, is far too small to detect.

09-12-2007, 06:55 PM	#8
Troll-Brain	ok, but at least we win ( i want mean the script will be a few more little) some bytes and it is faster to write ^^

09-13-2007, 04:04 PM

Troll-Brain

Double post

i've found how to test and i can say that using elseif is the best solution.
I'm happy to say : PitzerMike improve your code :P

By the way, really thx for widgetizer
PipeDream for grimoire.
Vexorian, for the jassnewgenpack,wc3mapoptimizer and of course Vjass.
And in general, all people who improve the war3 modding

Ps : i know that an integer becomes negative when you reach the max value, but in this example he had no time for this

09-13-2007, 04:49 PM	#10
Vexorian	well, if only strings had a < operator for lexicographical compare it could have been made much faster. I think that gamecache is faster in this case...

09-13-2007, 05:07 PM

#11

Troll-Brain

Quote:

Originally Posted by Vexorian

well, if only strings had a < operator for lexicographical compare it could have been made much faster.

I think that gamecache is faster in this case...

could you give an exemple plz, i don't uderstand what you said

09-13-2007, 05:10 PM	#12
Vexorian	Under the assumption < worked in strings: JASS: if (c<"a") then if (c<"P") then if (c<"D") then if (c<"B") then return 64 else return 65 endif //etc, etc etc. endif Later I said a single GetStoredInteger ought to be faster than 126 string comparisons. Although I forgot about gamecache's lame case insensitivity which will require an extra concatenation or at least native call.

09-13-2007, 05:26 PM	#13
Troll-Brain	Hmm maybe but why not use elseif instead of such many if/then/endif

09-13-2007, 06:56 PM

#14

moyack

I was thinking the same, but I realized that elseifs sometimes bugs. I'll take this opportunity to explain this bug that I recently found.

If you set a conditional in this way:

JASS:

function elseiftest takes nothing returns nothing
    if CondA then
        // CondA Actions
    elseif CondB then
        // CondB Actions
    elseif CondC then
        // CondC Actions
    elseif CondD then
        // CondD Actions
    //....
    elseif CondN then
        // CondN Actions
    endif
endfunction

If all the conditions before N are false, the CondN is evalueated, it doens't matter if CondN is false. The bad thing is that it happens sometimes, not always, and unfortunately I haven't been able to find the cause of this bug.

I'd like to know if this situation has happened to anyone.

09-13-2007, 07:33 PM	#15
Captain Griffen	Post the figures you got from any benchmarking.