Alienhead

Sounds like you have given this some tremendous thought. I for one, love the idea. Any plans for long term support on LE or is that pretty much done with? How portable will LE projects be to the Ultra Engine ?

Well the code base is rather large and as I stated earlier I had no idea where the problem was even occurring from. But good news! The only new media I added to the project this week was a skybox skin, I quickly pulled it out of the project and haven't see the error since. I do appreciate your willingness to help however.

Moving my entire main logic loop over to c++, I've stumbled upon this little error I've not seen before - mind you it's not crippling the game flow at all but I just hate seeing red in the console widow. Error states : Error: Texture can't be reloaded because it is currently bound to a buffer. As I mentioned, it's not causing any problems as far as I can tell with the game flow, no crashes or anything.. Just puzzles me because I want to fix it but I have little to nothing to go on.

I do however have one question that has arose since restructuring has been done - I have several prefabs that work on timers. They eventually Release() themselves once their lifespan has expired. Question being, how would one go about pre-loading these type of prefabs for later Instancing:: in code ? The only thing I have come up with so far would be to set the prefab script to inactive on cache-load, then when making instances of the prefab in real-time turn the instance copy to active. Not sure this would work however and it'll be a major overall to even test it out. Here's a example; this is a prefab consisting of a model and a script : function Script:Start() self.size = 80 self.active = false end function Script:UpdatePhysics() if self.active == true then self.size = self.size - 10 self.entity:SetScale(self.size, self.size, self.size) if self.size < 0 then self.entity:Release() ; end end end I turn the script off by setting self.active to false. Now I can load this prefab and hide it for later Instancing. Question being is once I instance this prefab and pop it into game play, how would I set the self.active to true from the calling function that originally creates the instance copy ... To further explain . flash= Prefab:Load("Prefabs/Effects/sprite_muzzleflash.pfb") -- this is called in my pre-mapload , it is the main copy I will instance off of. now later down the line in real time .... I may call this in some function - trail = tolua.cast(flash:Instance(),"Entity") Now since I have the active set to false in the main prefab, I need to set the active variable to true in the instanced version so it can do it's thing. How would I set a variable in that script from outside of that script ?

After following up on your advice, I restructured most my infrastructure to reuse the same prefabs/emitters etc.. Not only have I noticed a nice 20% increase in framerates and overall game smoothness - but I am no longer randomly crashing everywhere either. As far as the other error/warnings I was getting :: Possible reference count error for asset ::, this was directly caused by creating a CacheWorld() I was using. I created a seperate world on game startup : cache=World:Create() World:SetCurrent(cache) Then I loaded up tons of models I was using throughout all my maps. ( to keep them in memory for faster duplicating in the main code ). But somewhere over the development-span of this project the ref numbers got thrown off, once I restructured my main code to handle it's own caching and instancing I no longer needed 'cacheworld' so I ripped it out... Lo and behold - my assert warnings and errors vanished. So all is good in toon-land again, the show must go on !

--------------------------------------------------------------------------------------------------------------- function Monitor_Muzzleflashes() -- iterate: for i, m in pairs(mf) do if m.size <0 then m.enti:Release() m=nil table.remove( mf, i ) return end m.size = m.size - 10 m.enti:SetScale(m.size, m.size, m.size) end I've spent the last 3 hours moving my muzzle effects out of code an into a prefab'ed-scripted pivot created in the editor. It serves the exact same purpose but only no freeze up or shutdowns when I run it.

Something new - Possible reference count error for asset "c:/__ukr/materials/effects/default.mat" I havent done any manual reference asserting, not sure why this even came up. Haven't even used this mat, unless the engine loads it for some other purposes.

Never had a problem removing while looping, it's continuing to the next entry anyways.. Just to test it I removed the : if m.size <0 then m.ent:Release() m=nil table.remove( mf, i ) end from the function and it still randomly crashes. It seems to happen mostly after a physic collision detection.

I've got a strange thing happening, here's the code : -- iterate: for i, m in pairs(mf) do m.size = m.size - 5 m.ent:SetScale(m.size, m.size, m.size) if m.size <0 then m.ent:Release() m=nil table.remove( mf, i ) end end At random times, no reason whatsoever, this line m.ent:SetScale(m.size, m.size, m.size) will crash me to desktop ( no errors or anything in the debugger as far as errors or warnings), I've rem'd it out and no crashes, I then replaced the line with m.ent:Point(camera) and at random times for no reason it crashes the same way. The entity is not being touched anywhere else whatsoever. It's really been bothering me for a few days now. Not so SILIENT afterall.. Change the order of the function up a little bit and was able to get this in debug mode - "Asset map value is different" OK.. now.. WtF is that ? lol.

These screenshots make me want to design some form of modern day high-end mario bros. game.

I'll upload some media when I get back home, but I think I finally figured it out. I dumped the navmesh system and took the mobs off their character rigs and went with a straight forward waypoint node ( aStar type pathfinding ) and the frame rates picked WAY WAY up. I've kept the navmesh system in tact for player pets to follow and occasional boss spawns that chase the player around the map, but for the 100's of minions I released from character rigs and it's really improved the frame rates, its playable now on my good coding rig and even on a mid-range testing rig I got.. Thanks !

Youtube has been acting funny as of late with large uploads.. Anyways heres what I was trying to send up earlier today. \

Okay uploaded, it's processing, guess it'll post live in a few.

Okay, I've compiled a video - this video has a-LOT of animation stuff going on, birds, ants, butterflies, fish, mobs, players, wind .. its very animation heavy. What I did was remove the normal mob model I was using, which was 792 polys with 23 bones, and replaced it with a 200 poly 8 bone model just for testing.. I got 50 mobs running in the scene towards the end. It seems my frame rates were steady. Could this be the prior mob having 20+ bones in it ? Ahh hell, guess it would help if I upload the correct video .. one sec...

okay, I'm at work currently, but will send it up this evening, thanks !

release mode : 13 total animated entities on screen, less than 3000 poly, 21 bones, fps is at 39fps .

Surely not faster than faster than having no physics body to begin with ?. I have no player collision controller/body as the camera just floats in the air, its a town down tower defense style.

Heres the debug info -

The scene consists of a 256x256 terrain, no physics, very low poly meshes, characters have less than 20 bones.

No no no... No mass or physics of any kind is being used.. simple waypoint movement.. no navmesh. This simple scene right here is jerky on very decent hardware.

That's just it, I have ZERO physics working in this simulation, I'm building a top down tower-defense game and it requires no collisions. Just waypoint pathfinding. Also testing the simulation out on another PC, I noticed my camerapicks to map objects were way off, do I need to adjust entity picks from camera based on resolutions ?

I may of jumped the gun, while killing the debugger did help, I still have noticed that 10 or more animated models causes a huge jerkiness in the overall performance of the app. The test was ran on a blank map, 433 poly model with only 13 bones, no physics involved and I shut down the lighting. I'm starting to lean toward the idea that the desktop refresh rate the the virtual window aren't syncing.

Publish > Standalone ??

Problem solved, I've been hunting this for days now.. In case anyone else is running this setup - I suggest not. The cause was Tab-Nine intellisense engine combined with Lua debugger ( SubDebug ) running in Sublime 4. The reason it was so hard to find; the debugger stayed running as a service even when I shutdown Sublime, so I didnt look any further with sublime since I had shut it down.