Performance Optimization tips.

[JackFalcon]

Question about setEnabled(false). Is it better performance than visibility = 0 ?

And wondering what is the difference?

If 1 mesh is rotating 10 children, and 5 can be hidden - which is more performant: visibility or setEnabled(false);

I'll test it....

Lots of reading. 

Thanks!  

[brianzinn]

I would be very surprised if visibility = 0 has higher performance.  There are too many checks (ie: that break a for loop on mesh.isEnabled()) for disabled that are before visibility checks...  Interested to see any statistics.

[Hersir]

Hi @aFalcon

visibility = 0  will just not render mesh, but will do matrix calculations (position, rotation e.c), but if you set enabled false, it will not perform these calculations, so enabled false should be more performant.

[Dad72]

Hersir has right. setEnabled(false) no longer makes the mesh on the scene, but visibility=0 just makes the mesh invisible but still rendered as if it were visible somehow.

[JackFalcon]

Statistics -->great idea<--

Thank you setEnabled(true) insight!

TINY-UPDATES: 

- the code below worked well as first pass rough-in.

RESULTS: performance improvement achieved! (ding).

- Will try setEnabled... on parents

-  -> But had interesting PROBLEM: need parent setEnabled(true) AND children setEnabled(false)???<--

       anyway... : )

- Hello LOD! too.

- Probably date.now all three [hope] (BZ thx).

reloadLevel = function(){
    for(var i=0; i<edges.length; i++){ edges[i].visibility = 1; }
    for(var i=0; i<orbs.length; i++){ orbs[i].visibility = 1; }
    for(var i=0; i<boxes.length; i++){ boxes[i].visibility = 1; }
    for (var i=0; i<tiles.length; i++){ tiles[i].visibility = 1; }
}

unloadLevel = function(){
    for(var i=0; i<edges.length; i++){ edges[i].visibility = 0; }
    for(var i=0; i<orbs.length; i++){ orbs[i].visibility = 0; }
    for(var i=0; i<boxes.length; i++){ boxes[i].visibility = 0; }
    for (var i=0; i<tiles.length; i++){ tiles[i].visibility = 0; }
}

 

[JackFalcon]

Ok,

Put in TIMESTAMPS ... gotta go.

Thanks to who'ever made dude.

Here is adapted timestamp utility, to easily measure performance anywhere:

//TIMER
var timeStamp = 0;
var currentTimeStamp = 0;
var timeReport = function(testname){
    console.log('TIMESTART: '+testname);
    if(!timeStamp){ timeStamp = new Date().getTime(); }
    return function (testname) {
        currentTimeStamp = new Date().getTime();
        var timeDelta = currentTimeStamp - timeStamp;
        console.log('TIMESTOP: '+testname+' = '+timeDelta);
        timeStamp = null;
    }
}
//EXAMPLE-USAGE-.
// var debugTimeStop = timeReport('START Time Test');
// debugTimeStop('END Time Test');

Statistics pending...

[JackFalcon]

rudimentary stats... 

TEST DESCRIPTION: rotating parent with gradual increasing number of children(30,60,90) for each rotation, step-interpolation = 0.02, on a closed-circuit track.

BASELINE   NO-PERF-OPS   3 CONTROL-TESTS:

TIMESTART: START-ANIMATION
TIMESTOP: END-ORBIT = 1867
TIMESTART: START-ANIMATION
TIMESTOP: END-ORBIT = 4294
TIMESTART: START-ANIMATION
TIMESTOP: END-ORBIT = 6724

 

OP-TEST-1  visibility = 0    3 COMPARISON-TESTS:

TIMESTART: START-ANIMATION
TIMESTOP: END-ORBIT = 1937
TIMESTOP: END-ORBIT = 4237
TIMESTOP: END-ORBIT = 4679

OP-TEST-1 RESULTS:

Within variance, visibility = 0 for hidden mesh, reduces orbit time ~2 seconds on third rotation (HOO-RAY).

QUESTION: which is better visibility or setEnabled(false)?

Bird should read source code, or...

TEST: replace vis =0 to  setEnable(false) in the same loops...

TIMESTART: START-ANIMATION
TIMESTOP: END-ORBIT = 1917
TIMESTOP: END-ORBIT = 4227
TIMESTOP: END-ORBIT = 4778
 

RESULTS:

Interesting - visibility is slightly faster (within variance). 

Didn't expect that. Bird should read source code.

NOTE: couldn't use setElement(false) at parent level because in context need parent to remain visible, and only children hidden.

NEXT-STEPS: 

Not going to do LOD test, sticking with simple visibility op for the moment.

Instances not applicable.

This is solved. Maybe more PERFOPS later.

Hope this helps someone.