GenkiGenga cheers mate, thanks for your answer. Yeah I think my problems are mainly code issues. Me and my loops! Im placing triggerOnces everywhere I see fit, and so on. My graphic assets are all png8, which I hope makes it all faster, as well as lighter.

Glad to say I made some improvement. I had some "proximity checkboxes" placed in a parallax 0 layer, which was no good for C2 collision cell sysytem (i was gettin like thousands of checks per tick -- now I'm around 20-40!). Game is quite playable now , though still slower than intended (some 40 fps). My objective is not less than the damn 60fps on phone. Is that a realistic goal for a big layout game? Is that... really necessary, too? I mean, do commercial mobile games (not AAA games, just fine games) stick to the 60...? Right I know you can't get universal 60fps , but -- should I definitely try to achieve them in my own (mid-end) phone? Guess wont be an absolute answer, but what you guys think?

newt means actually generate random dungeons... cos what you're pointing is not real, procedural generation, but some random assign of premade rooms... and i say you stick to that, unless you wanna go into heavier coding.

[quote:31sg2liu]But maybe these two things are simpler than i think they are.

I'd say they are. Anyway the sprite could work as a "template" for the room, and onCreated it could spawn its own walls, wandering monsters or whatever. I dunno. There's always many ways...

I'm trying hard to optimize my (goddamn) game, as it's running at sluggish 20 fps on phones. I know Scirra's policy is "measure it yourself" so that's what I'm trying to do, but I find it (boring and fustrating as fu--) difficult for a number of reasons. No, only one reason actually: ignorance. Mine.

So I open the debugger, and there I'm basically looking (Watching) at:

FPS obviously,

Est. CPU utilisation,

Collision checks per sec,

Poly check per sec.

I also have me a big tileset, of which I'm overseeing the Draw and Collision rects... and then the global CPU performance on the Profile Tab, to see if any loop has gone wild.

Obvious question: is there anything else I should be keeping an eye on?

(Not like I'm like, blind to anything else, but yknow).

Also. I feel like I lack references, beyond the 60 fps. I'm aware my question here might just miss some abstract technical element at its base that make it unanswerable or something (i really know but very little about devices), but i'm gonna try anyway: is there any reference values I can hang to for things as... well, as those listed above-- CPU, collisions, polygons? Which range of values are mobile-safe for polygon checks, for example?

I guess (and it's a guess) that those values may depend heavily on the machine you are working on... so lets say for your average, win10 good ol laptop? -- just like mine? ;P

Thank u guys

You could spawn them rooms on the go. Supposing, for claritys sake, these rooms are simple sprites. You create an image point in each door. Then as you approach a door (but before it's seen on screen), you spawn room number random(10). You'd still need to figure something to make both rooms' door connect properly, avoid room repetition, etc.

Very cool game! unfortunately I reach only 50 fps. It might be due to the filter effects tho. Or might no -- ill be damned if i ever understand how performance actually works...

Thanks guys. To be honest Im not that sure anymore that my problem is in the tileMap itself (as newt pointed, if there's no collision at all it shouldnt give me collision trouble). The draw calls could be, but it dosnt really looks like that, although Im no even sure I'm reading the debugger properly...

My new fav suspect is the "solid" invisible platforms (nearly 2000 at the moment, but there will be many more). PassThrus too, tough there's not as many as solids. Monsters and items are created on proximity and destroyed at a certain distance, there shouldnt be problems with that. The thing is, when i try to check the solid platforms for collisions their bare numbers seem to slow everything down, so I think I'm havin a read on that blog entry that Colludium mentioned and see if it fix me.

clarkkentnicdao well it would be a 25600 x 14400 px sprite! Many sprites, could be... I was pretty sure my tilemap option was better and more efficient, but honestly Im no that sure any more -_-

thank again for yr help brothers

yeah, nevermind the loops! im not that good at coding

newt thanks mate, i think i see now. Guess is something similar to what i am doing with enemies and other stuff: they only became "active" when they overlap a certain "proximity box".

So, with no collision or overlappin event or behaviours on the tilemap there should be no performance impact? Still im getting collision rects (rectangles, guess its each tile?) on the debugger.

I think you are right, the problem may be elsewhere. I'm not sure I'm readig the debbuger correctly too.

clarkkentnicdao not tiledBG but a tileMap. But yeah, it makes the bulk of the scenery and covers the whole layout, with separated sprites here and there to represent animated or interactive bits.

Ninja by Daylight is it? I dont know dude it works alright for me.

...wot? ·_·

thanks for the answer newt , but i don't really think Im getting it... What you mean, a variable comparison?

I'm not saying individual tiles, but the whole object thing -- make the whole tilemap into a merely decorative thing (as platforms an jumpthrus and stuff are implemented separatedly).

Yeah Im having performance issues. Getting around 10-15k collision checks per second. Im not savvy on the art of debuggin, but i think thats a lot. Not even 100% sure it's the tilemap -- but me tilemap is 360,000 tiles big (25600 x 14400 px), tiles at 32px, and uses currently 68982 collision rects, which i also think is quite a lot. On the bright side, I'm using png-8... and, again, only need the tilemap itself for decorative purposes.

I'm beginning to fear i should have used many small tilemaps instead of this monster. Well i could split mine into smaller ones, but this would be rather shitty work (and just the kind of thing you spend hours doing and tweaking and fixing just to realize on testing that performance blues is still there), so i'm checking other possibilities first.

There's always the possibility, or even probability, that I'm doing other things wrong too. I'm doing a big sandbox, which is always tricky on performance, but (i think) I'm properly disabling behaviours and stuff of faraway objects so they don't mess with frameflow... anyway, this should be stuff for another topic -- if this one finally doesn't ease my worried mind ;P

bitballoon.com does it beautifully

Yeah well, it's hard to tell without a capx to try and see, but hey why are you stopping the animation after each frame change -- it's easier and safer just to set the animation speed (first image, under Animation 'Default' properties) to zero, and then you can just set the frames at ease.

Im not sure if that would fix you tho, you may need to use some System>Loops>ForEach conditions here or there...