Graphic System

03-17-2007, 01:18 PM

The Graphic System is a library that makes certain graphical aspects in Warcraft 3 much easier.

Version: 1.3
Last updated: March 25 - 2007

Features
- Easily create primitive and more advanced static shapes such as circles, lines, spirals and more.
- Flexibility: manipulate the parameters of the shapes created in many ways.
- Geometrical checks
- Dynamic effect birth & death: create static objects with a slight delay between units composing them and destroy them in the same manner.
- Timed effects.
- Dynamic and static destruction of the effects.

Promising features to come
- Dynamic effects: orbiting around units/a certain point, dynamic wheels
- Lightning pools: effects with lightnings
- More static effects (eg: stars)

I expect criticism to fix possible leaks, flaws and problems. Suggestions are heartily welcome as this system is made for the modding community for great usefulness and easy implementability. Expect new versions.

History:

Version 1.0 - First version of the system with the basic static shapes (circle,arc,line,spiral,hypercycloid).
- Basic geometric checks.
Version 1.1 - Dynamic birth and death shapes included.
- Radius reduction function added for safety reasons.
Version 1.2 - GS_DestroyEffectDynamic() function added.
- Timed versions of both static and dynamic (fixed) shapes added.
Version 1.21 - Epicycloid shape added and fixed a small bug with hypocycloids.
Version 1.3 - Rose shape included
- Fixed major bug which caused imperfect (and most of the times incomplete) shapes.

Credits: Thanks to Vexorian for the dummy unit model.

~Daelin

Attached Images

GraphicSystem.jpg (171.3 KB)

Attached Files

Graphic System.w3x (57.6 KB)

03-17-2007, 01:40 PM

Rising_Dusk

JASS:

function GS_Sin takes real angle returns real
   local integer k
   local real kr

   set angle = GS_ReduceAngle(angle)
   set kr = (angle-R2I(angle))*GS_Position()
   set k = R2I(kr)
   if (kr>(k+0.5)) then
      set k = k + 1
   endif

   set angle = R2I(angle)+I2R(k)/GS_Position()
   return udg_LSine[R2I(angle*GS_Position())]
endfunction

I just have to ask.
How is calling 4 wrapper functions (Each using at least 15 LoC) in a function that in itself has 13 LoC and is slower because it's a wrapper going to be faster than just calling Sin()?

While I realize Sin() is slow, I'll be damned if it's that slow.
Your Cos() replacement is negligibly better.

If you've done tests on it and somehow it is faster, then please cast aside my ignorance.
But seeing that just made lil' red alert lights go off in my head.

And erm...
Maybe it's ironic here, but your "GS_CircleQuick" function uses TSAs en mass inside of a loop.
That would be pretty... Slow, and inaccurate on BNet.
Timers would be far superior.

I really like the PART 5 stuff for geometric tests.
Some of those checks really aren't intuitive and in themselves are pretty useful to have 'round.

But yeah, looks like it would be useful for some people.

03-17-2007, 01:41 PM

Vexorian

The idea to use precalc for trig functions appeared in the forums one day. When I benchmarked stuff, native Sin was faster than a non-native function call... Of course, the one to blame would be blizz for making function calls so slow, but there you go.

Gamecache is not exactly the best idea. I would have done something like this:

JASS:

globals
   effect array GS_linked1
   effect array GS_linked2
endglobals

function GS_H2I takes handle h returns integer
   return h
   return 0
endfunction

function GS_LinkEffect takes handle pivot, effect which returns nothing
 local integer i=GS_H2I(pivot)-0x100000
    if (i>=16382) then
        call StoreInteger(udg_GSCache, "GS_Effect",I2S(i) , GS_H2I(which))
        //swapped it, this should be faster and take less memory since it will have a main hash and many values in that hash,
        // instead of a lot of hashes one value each
    elseif (i>=8191) then
        set GS_linked2[i-8191]=which
    else
        set GS_linked2[i]=which
    endif
endfunction

function GS_GetEffect takes handle pivot returns effect
 local integer i=GS_H2I(pivot)-0x100000
    if (i>=16382) then
        return GetStoredInteger(udg_GSCache, "GS_Effect",I2S(i) )
    elseif (i>=8191) then
        return GS_linked2[i-8191]
    endif
        return GS_linked2[i]
endfunction

edit: Darn dusk

03-17-2007, 01:46 PM	#4
Daelin	The TSA was mistakenly left there... I forgot it there. Sorry bout that. And indeed... I intuitively found that the Sin would be faster as it doesn't actually calculate the value. It's time to switch to Grimoire I guess. Edit: Just saw Vex's post. What's wrong with gamecache anyway? ~Daelin

03-17-2007, 02:00 PM	#5
Vexorian	After seeing the whole graphics system and the precalc trig functions I guessed you were pursuing for speed?

03-17-2007, 03:23 PM	#6
Oglog	Wow looks very nice. And easy to understand and implement too. You should definatle contiue to develop it as i'm sure it will become very popular for making effects. Maybee you should check out http://www-groups.dcs.st-and.ac.uk/~...es/Curves.html if you haven't seen it allready.

03-17-2007, 05:35 PM

Daelin

Oh yes Vex... about that, I just wanted to make certain effects as fast as possible (for example the Cos and Sin stuff).

I have a request if someone could nicely help me with it. Would anyone offer to test whether Sin is actually slower than my GS_SinQuick? Because of it isn't, I'll just drop those functions and focus on other aspects of the system.

Thanks Oglog... I was planning to implement other curves too. I remember some "flower" made out of multiple cusps I think. I may try to implement something like that too. As you find it useful I'm sure many others will, so what the heck? Time to continue with developing.

~Daelin

03-17-2007, 07:31 PM	#9
Oglog	Maybee a Sinusoidal where p = 2.5?

03-17-2007, 07:55 PM	#10
weeg2	oooooh soo preety

03-17-2007, 08:40 PM	#11
Av3n	Isn't it iFraNe dummy model? -Av3n

03-17-2007, 10:00 PM	#12
Vexorian	It is my dummy model, I used INFraNe's model coordinates for animations the rest is mine (my version takes less space...)

03-17-2007, 10:52 PM

#13

Vexorian

Sin vs. DoNothing, the benchmark:

JASS:

function Trig_Melee_Initialization_Actions takes nothing returns nothing
 local integer sw=StopWatchCreate()
 local real t0
 local real t1
 local real t2
 local real t3
 local integer i

call TriggerSleepAction(5.)


    set t0=StopWatchMark(sw)
    set i=1000
    loop
        exitwhen i==0
        call Sin(GetRandomReal(0,7.283185))
        set i=i-1
    endloop
    set t1=StopWatchMark(sw)
    set i=1000

    loop
        exitwhen i==0
        call GetRandomReal(0,7.283185)
        call DoNothing()
        set i=i-1
    endloop
    set t2=StopWatchMark(sw)


    call StopWatchDestroy(sw)


call BJDebugMsg( "DoNothing: "+R2S(t2-t1)+" vs Sin: "+R2S(t1-t0) )
endfunction

"DoNothing: 0.122 vs Sin: 0.092"

03-17-2007, 10:54 PM	#14
Av3n	So u used iFraNe animations in other words? Lucky yr admin double poster -Av3n

03-18-2007, 02:02 AM	#15
Deathlust	That is hawt... +rep... Even if it doesn't work correctly it's still hawt!