TiAm's Forum Posts

  • My understanding is that WebGL effects can be tricky on mobile because some mobile GPU's have odd architectures whose shortcomings can cause effects to fail entirely.

    For example, on my android phone, the noise effect fails due to the way my Adreno GPU handles the sin() operation used to generate the noise effect, as per this article:

    http://byteblacksmith.com/improvements- ... gl-es-2-0/

    I know I've also heard of seemingly cheap effects grinding games to a halt on mobile because the particular GPU has trouble handling certain mathematical operations or data types in an efficient way.

    Then again, you might be hitting a simple fillrate issue. Try a lower resolution and see if that clears things up.

    In my case, while the noise function fails, most other effects work fine. I have an Adreno 305, and I continue to be amazed that I can pile on multiple webGL effects without tanking the FPS. Yet, I'm CPU limited: a couple dozen physics objects brings me sub-30fps framerates.

    That's mobile: you have to figure out what the bottlenecks are, because they simply aren't consistent. It's best to try and build tests into your own game, enabling/disabling effects depending on what the FPS is doing.

    I've been thinking lately that it should be possible to test effect compatibility on mobile by having startup tests where effects were applied to test sprites, sprites captured to a paster object or via the CanvasSnapshot function, then processed as an array to check if the effect is actually working.

    So, see if a white sprite with the noise effect has mostly uniform pixel values (effect is failing), or if all the pixels seem different from each other (effect is working). For HSL, transform a small sprite, and check that the captured color is the same or close to what it should be. Etc...

  • A directional blur with 4 attributes:

      Samples: Angle: Length: Z-Order: on top/underneath

    With this effect we could accomplish a reasonably cheap motion blur effect by setting blur params with object angle of motion (angle) and speed (length/samples).

    Could also be combined with the expensive but awesome looking multi-sample approach here (which would look way better with a proper directional blur effect):

  • The best thing to do is just look at other plugins, and ask on the forum, where, unlike some forums, you'll almost certainly get an answer.

    The SDK documentation has never been that extensive, and somehow quite a few people (including myself) have managed to put a plugin together. Most plugins are freestanding from the engine code, but solid is one exception.

    Unlike Unity, there really isn't an engine SDK, which is unlikely to change for C2, but seems to be on the radar for C3.

    A lot of people have expressed their frustrations with the limitations of the solid behavior. I'm going to hazard a guess that Ashley is aware of this.

  • I've heard rumblings of SIMD coming to JS for awhile now, but I just now got around to watching this talk by Moh Haghighat from Intel. As I only had a basic understanding of SIMD (and vectorization in general), I found this a really enlightening peek into the tech.

    Youtube:

    Subscribe to Construct videos now

    In short, SIMD.js will not necessarily require custom code on the part of javascript programmers, as future JIT engines will automatically optimize incoming code to take advantage of SIMD where applicable. Of course, explicit use of SIMD functionality is coming first, as JIT integration will take more time.

    Also, many mobile platforms already support SIMD via NEON, something I had no idea about.

  • Glad to see some slow recovery from the disaster that was chromium v38/nw 0.11.X

    10.5 still works better, but 0.12.x is certainly a viable option, unlike 0.11. Here's hoping it zooms ahead by stable.

    XP support...well, it had a damn good run, but honestly, I see this as a positive step, much for the reasons you mentioned (users on XP are liable to be a support issue anyway). Win 7 is the best OS Microsoft has made so far, and hopefully 10 will be another step forward after the mis-handling of Win8.

    BTW, Ashley, great idea to 'alpha test' a new version of node this way prior to update. Hope to see this become the standard approach moving forward.

  • First off, I'm still seeing the HSL glitches mentioned previous by other users. Only seeing it on certain colors. It persists after export as well. My graphics are intel HD4000, so that might be part of the problem. Anyway, here's an example of what I am seeing:

    [attachment=0:z32q4vn8][/attachment:z32q4vn8]

    While I'm here, I'd just like to say this about Stuffgen:

    As a designer first/artist last, this is insanely helpful for me when prototyping, as I tend to get hung up on art and lose my flow sometimes. I can always get something passable, but since I'm just not very efficient in that area, I lose precious time worrying about dev art details that are often meaningless in the context of a brainstorming session.

    Stuffgen is brilliant for this, as I can work up a bunch of temp sprites in no time, and get back into events and programming, which is generally where my real ideas are anyway. With the way things are progressing, I could even see using some of these sprites in a real game.

    So...thanks — For preserving my brain. In a totally cool, non-mad-scientist sort of way.

  • Thanks for contributing this Phobos002. <img src="{SMILIES_PATH}/icon_e_smile.gif" alt=":)" title="Smile">

    I modified your last capx to add a velocity dependent blurring effect that makes this look even better in action, and is useable down to 4 samples for slower movement (lower settings blur more, which sacrifices accuracy, but still looks smooth in motion). Here it is:

    CAPX: https://www.dropbox.com/s/9j0q36mne5x56 ... .capx?dl=0

    Be aware, this is a much hungrier effect GPU-wise (blurring is being applied to each moblur sample individually as a webgl effect).

    Still, at 12 samples it's smooth on my HD4000, and it could probably be used in a real game w/ a dedicated GPU. As an incidental effect, it could be used just about anywhere.

    Could make the effect cheaper/better looking with a proper directional blur effect, but I'm not aware of one for C2. Also, could probably batch moblur sprites on their own layer(s) for considerably cheaper blurring at high samples, but that would require a separate layer for each set of sprites (not un-doable, but a PITA to be sure).

  • This sounds like a good solution, though I haven't had a chance to try it yet. Thanks, Joannesalfa.

  • As far as I can tell, I'm no longer seeing the flipped image with the last example you posted.

    Glad to help with the testing; thanks again for this great plugin.

  • I'm sure I'm being a total dunce here, but I can't figure out how to call up the internal android keyboard and fetch input from it. I assumed the keyboard object would do this, but it says specifically that it won't.

    Help?

  • For any serious projects, I stick to the stable versions. However, I keep a separate install of the latest beta version too, in case I want to try out a new feature or open a newer capx file.

    C2 is pretty much portable, though you wouldn't know it from the installer...you have to be careful when installing beta's or you'll uninstall/overwrite your stable version.

  • Try Construct 3

    Develop games in your browser. Powerful, performant & highly capable.

    Try Now Construct 3 users don't see these ads

  • TiAm

    Try this:

    http://tinyurl.com/neye756

    tap to paste the canvas.

    Does it flip?

    No, it seems to be working correctly in all versions of chrome (stable and canary), and also in Opera. FF/IE don't work, but FF was inconsistant before, and IE never worked.

    Canary was updated to v42, but I'm still seeing the backwards image when I preview from my own project using the Paster plugin, as well as from your example capx. Your newest example on dropbox seems to work fine, though.

    I tried to export my project as an HTML5 page to see if that made any difference; it seems to give the same result as preview in all browsers. I PM'd you a link to this export in case you would be interested in testing w/ Canary on your system.

  • Update:

    Tried an XDK export with v7 stable, and the paste from canvas effect seems to work aces.

  • R0J0hound

    I'm getting generally inconsistent results, but it seems like the newest update is actually worse in preview: on android I'm now getting the upside-down/mirrored effect I was seeing on Canary/FF. On desktop, I don't see any difference.

    I reverted back to the previous version of paster and the android preview seems fine again. I tried updating/reverting the plugin a few times: once it seemed to be working correctly, a couple times nothing was pasted at all, and most of the time I got the backwards paste.

    Going to try an XDK export and see where that goes.

    > Somebody

    > This may be useful for you for saving part of the window at any size image you want.

    >

    Wow! This is simply beastly - I suspect it also enables a feature I was kinda going to mention as a slim possibility - supersampling (in a way) - i.e. punching a higher res image into a lover res canvas and thus improving the quality of the result.

    Speaking of which - can I set the resolution to beyond object size? Would the pasted objects then be, for example, pasted at a higher res. And can I change the res and thus downsize that image before saving?

    Wow, I hadn't thought of that...supersampling would be awesome for more powerful PC's, and this would be an easy way of doing it.

    I've noticed before that you can force downsampling in chrome by zooming out; if high-dpi and high quality scaling are enabled, chrome will render to a higher rez, then downsample in the window. It really smooths out some of the hard edges induced by some blend effects.

    Being able to do that in an actual game...seriously amazing.

  • Okay, this is just freaking odd.

    To sum up: I have a 3570k w/ HD4000 graphics. This is only happening in chrome canary and firefox. Here is the tally (triple checked this time):

    Android:

    The latest chrome, v39, works fine. Firefox doesn't capture anything, just gives me a blank paster object (big surprise).

    PC - Chrome:

    Chrome Stable v39 works fine. Unbeknownst to me, I was testing in Canary before (it comes up when I select chrome in C2).

    Canary v41 is doing the odd mirror and flip trick, plus I can see thru the paster object slightly, as if it had an opacity of ~96 or so.

    PC - Other:

    Firefox does the exact same thing canary does...sometimes. Other times, the paster object remains blank. Odd, considering Firefox is a separate code base.

    IE 11 doesn't do anything; the paster object remains blank.

    Node-Webkit 0.10.5 works fine as well. Not sure about the newer versions (though who would want to use them, anyway).

    See any of this on your end?