I need your help to figure out Airscape's performance issues

Discussion and feedback on Construct 2

Post » Tue Apr 07, 2015 9:14 am

It may be worth it to check also pix.js on a non intel gpu to see if the ratio is still as big, and also to test it with moving objects (note I am on a tablet and the test you brought does not respond to touches so I do not know if those are moving), as there may be a optimisation for non moving objects to prevent rendering them again if not needed on a smaller level than the everything or nothing moves of C2.

Still an intel problem if their gpus reacts differently of course, but if there is something that can benefits everything, might as well do it. It would be interesting to know what they are doing differently (worst case: not using functions because of the intel issue), if it is for a good reason, might be nice.
Game design is all about decomposing the core of your game so it becomes simple instructions.
B
43
S
18
G
18
Posts: 2,106
Reputation: 16,436

Post » Tue Apr 07, 2015 9:58 am

@Aphrodite

As for moving vs nonmoving objects: If you want to test moving objects, use Bunnymark. Also, as for an optimization to not render nonmoving objects - I don't think it's that simple - as far as I can see, there really is no point at all in such an optimization because it would be rendered useless 99% of the time, i.e. with scrolling, or opacity, or anything at all being behind the object...

The problem I'm going to face is Intel telling me it's a C2 problem, since 'it works fine in every engine other than C2!' And others will say it's still an Intel problem, 'because all other GPU's work fine.' We're never going to get anywhere.

Also, hmm, maybe I should make a new topic or something because this really isn't specific to Airscape anymore... even though for some reason nobody else has ever noticed this problem? agh...
B
92
S
30
G
24
Posts: 3,189
Reputation: 32,570

Post » Tue Apr 07, 2015 10:36 am

@sqiddster I understand, and I agree that until we know for sure, both parties should try to improve the situation, also I asked for pix.js on other GPUs as, if it is still much faster than C2, then we would know it is an intel issue BUT that there would be a sorta large room for improvement on the C2 side possibly, which would help everyone I think.
Game design is all about decomposing the core of your game so it becomes simple instructions.
B
43
S
18
G
18
Posts: 2,106
Reputation: 16,436

Post » Tue Apr 07, 2015 10:41 am

Surprisingly on my nvidia 780 card (newest drivers) with scirra renderperf (webGL version) I get ~112.000 before frames start to drop. On your pixi.js version I get only ~89.000

Don't have any Intel gpu to test thought :/
ImageImageImageImage
B
154
S
61
G
40
Posts: 2,574
Reputation: 33,783

Post » Tue Apr 07, 2015 10:44 am

@sqiddster btw. could you in spare time upgrade your benchmark with touch controls? I would really like to see a difference on mobiles between them (scirra and pixi).
ImageImageImageImage
B
154
S
61
G
40
Posts: 2,574
Reputation: 33,783

Post » Tue Apr 07, 2015 11:18 am

@sqiddster

Iw tryed that web demo on my old notebook and it runs at 15-20 fps on intel integrated graphics
i dont have steam there so i did not test steam demo.
one thing iw noticed... on notebook iw somehow managed to squeze squidster through wall and into the ocean... got lost and restarted level.
that might point to some issues with fps and delta time? couse i assume thats not supposed to happen!?
B
26
S
11
G
2
Posts: 669
Reputation: 5,038

Post » Tue Apr 07, 2015 12:36 pm

@irina yeah, because there's currently no way to cap dt to a lower value (an issue I've brought up a few times), glitches like that will happen all the time at lower framerates.

@shinkan I tried to add touch but it doesn't seem to be working, any js people able to tell me what I'm doing wrong?
B
92
S
30
G
24
Posts: 3,189
Reputation: 32,570

Post » Tue Apr 07, 2015 1:29 pm

I ran through your demo on my computer. I got a stutter briefly when leaving the menus, but never anywhere else. My FPS stayed at 60 for the most part. At times it would shift back and forth between 59 and 61 but nothing that was able to be perceived. DT difference was 2 or 3 pretty consistently. I'm posting a shot of the plotter just to show the 59/61, though I really don't think that's an issue. I had C2 with my game and Photoshop running in the background, though I don't think those would make a difference in this case.

Specs: i7-4770 @ 3.40 GHz, 32.0 GB ram, NVIDIA GeForce GT 640 with 4 GB vram

Image
B
96
S
34
G
16
Posts: 940
Reputation: 15,905

Post » Sun May 10, 2015 1:52 pm

On Nereus and Orbit (before I moved Orbit to Unity) I would experience exactly the same issue where shrinking the game window would improve frame rate performance. I fixed it on Nereus by limiting the number of objects in my scene. I would randomly generate an asteroid field, and found that if I reduced the number of objects in the level to <1000 performance was rock solid, anything like 1500 to 2000 and it would really start to chug on my laptop BUT weirdly, shrinking the window would bring the frame rate back right up.

Another prototype I worked on had a massive tilemap, I'm talking a layout size of 8960, 5040 with a tilemap covering a lot of it made up of 16x16 tiles. Again, bad performance at full screen, solid 60 when reducing the window size, reducing the internal rendering resolution and all that didn't seem to make any difference.

On my desktop (which is very powerful) I get a solid 60 on the Airscape demo fullscreen. I can't test my integrated chip laptop right now will edit post with results later.
B
40
S
12
G
8
Posts: 590
Reputation: 7,153

Post » Mon May 11, 2015 2:31 am

Hi all, I have been following this thread but until now have not performed the tests - finally got around to it today, here are my results.

Before dropping below 60fps:

C2 WebGL:
    about 24000 objects
C2 Canvas2D:
    about 4000 objects
Pixi.js:
    about 18000 objects

All tests at 1920x1080 res.
Edit: Window size seams to make no difference.

Test machine specs:
Intel i3-4160 CPU @ 3.20GHz
8 GB RAM
GeForce GTX 750
Win 8.1
If it can be done, someone on the web will show you how to do it!

CDASI Games Mentality Break Splat-a-bug FlapFleet Challenge
B
30
S
10
G
6
Posts: 335
Reputation: 6,300

PreviousNext

Return to Construct 2 General

Who is online

Users browsing this forum: Google [Bot], Lancifer and 5 guests