Ashley

just tried this:

set text to renderer, webgl on, then exported awesomium.

the hd4670 with catalyst 10.3 still says canvas2d (of course thats what the conclusion was before, because the fps behaved like a crippled canvas2d without vsync)

to compare, r106 is the FIRST version that works with webgl enabled on the x1950. before it didnt start with webgl but worked very buggy with canvas2d, somehow similar to the other setups only worse (picture bug).

now that the x1950 is working - there is is about twice the difference in fps when i enable webgl and export with awesomium.

without webgl enabled i get half the fps with awesomium but still really playable and working about 50-60fps (like chrome with canvas2d, no bugs, no shaders)

if i had to guess - something about detection, blacklisting or initialising, maybe smartshader support detection ?

"Awesomium exports from r106 should force WebGL on all the time regardless of the graphics card."

hehe i would prefer if the option of canvas2d stays - both awesomium builds work pretty good on the x1950. (beside from the other problem with startup and crashing on exit)

i read in past that in some countries paypal has limitations.

if it worked until now, search for a policy change update on the net (or maybe a terms and conditions mail in your box) and write them a mail in case you dont find anything.

gl

have tried a game on 3 different pcs with windows xp.

as i dont think there is a major difference why the games are working or not working there or in win7 i post it in this thread.

r106 is the best version so far with the awesomium exporter !

but like i read in the forum and can confirm, its still missing some

compatability and stabiliy.

1. on every tried computer i have to start the .exe twice before it starts the game.

2. on one pc it crashes on exit with an error message

3.

ati x1950 (catalyst 10.2 latest for that card) got for the first time with the exporter full speed like mentioned in other threads, with and without webgl ! so thanks for that, its a huge difference in fps

ati 4679 (catalyst 10.3, not latest driver for that card) has not webgl support and the mentioned slower than chrome performance.

intel extreme integrated (latest driver) no webgl and slower than chrome.

the vsync problem is there with the ati 4670 and intel extreme. so i would say its somehow related. when vsync is not working its looking

like playing a console game on tv with vsync disabled, if someone knows this effect of unregular updating and lagging picture. not like disabling vsync in pc games when getting more fps and only seeing the picture update.

anyway r106 has so far the best support of hardware, would be great if the next stable version could expand on this

error: "Procedure entry point DecodePointer could not be located in the dynamic link library KERNEL32.dll."

dont know if it helps, but this is the same error someone gets when trying to start c2 on xp service pack1.

so maybe it has to do with installed or not installed service packs on win7 x64 ?

normaly what it says is that it tries to access pointers which are not present in your kernel. in many cases this would be an older kernel version.

--

Window wrapper

Error: Missing installer and application has additional modules that are needed

--

winxp

"@makedit - if you are trying to maximise performance, you should enable WebGL. It's over twice as fast in Chrome and Firefox. Why have you disabled it?"

the last several times ive looked upon the webgl support of browsers i read something around 50percent.

codeflow.org/entries/2012/apr/25/webgl-statistics-and-the-state-of-webgl-html5

no support in ie, bad support in opera and safari. besides that i calculated even if the webgl support should be over 50 in the browsers people still have to meet the driver blacklist which is not really satisfying when ive tried it on several pcs (low to medium hard and software combinations - which i would call normal office pcs nowadays - but which all should run 99percent of desktop d3d and opengl games, and canvas2d too with the right browser)

so while i personaly like the idea of gl in a browser, i have tested alot of canvas2d prototypes and think these can be run at 30fps with medium graphic settings like a few parallax layers and 800x480 upto 1024x600 fairly well. thats why i guess im trading some gfx for more browser consistency.

so why do i switch wegbl off? there are some combinations ive encountered where webgl works, but is considerably slower than canvas2d and makes a game stuttering alot. but when a game is programmed within the limits of canvas2d of course theres no aaa gfx but it has a higher chance of working like it should. and people cannot mess it up when changing their browser settings, their hardware or update their browsers or drivers.

k for the problem i think ive nailed it down doing benchmarks in many browserversions:

Platform 9 (jump-thru).capx from the example folder

switched fullscreen off,

webgl off,

every tick - text to fps,

r84

opera 10.63      77fps

opera 11.50     79fps

opera 11.62     80fps

firefox 8     51fps

firefox 11     51fps

firefox 1401     54-56fps

chrome 13      58-60fps vsync disable didnt work

chrome 20     87-89fps

safari 517     78fps

r85

opera 10.63      37fps

opera 11.50     37fps

opera 11.62     35fps

firefox 8      53fps

firefox 11     53fps

firefox 1401     53-55fps

chrome 13      58-60fps vsync disable didnt work

chrome 20     87-89fps

safari 517     39fps

it seems the change mainly occurred in opera and safari. i cannot test ie9 because i have no win7 installation atm.

first, congratulations on r100 !

installed r100 to test if anything has changed in performance in canvas 2d. after seeing it has the same fps as r95 i got the idea to test which version broke the high fps count in canvas2d. tried r95,r90,r86,r85 and r84.

result: the change from r84 to r85 is where the fps got a serious hit

for me on several pcs. most noticeable in opera (half fps count).

i recognise c2 was overall more focused on webgl and cutting edge html5 technology lately. personally i would like to see a little shift more to stability and performance (eg canvas2d) as i dont see the consumer numbers in webgl. of course thats arguable and maybe preference at which point most people have good enough hard and software.

besides that, the same applies to the .exe exporter. imo it would be nice if the exe exporter runs on more hard and software combinations, maybe even including the option to use canvas2d if better webgl support is technically not possible or too much work.

also i recommend some minimum requirements page, where it list the minimum requirements for html5, canvas2d, webgl, audio, browser versions, exe export, mobiles and os versions... .

so that devs can expect which oldest configurations has been tested and are supported! now and in future. something like the minimum requirements on pc box sales.

any opinions welcomed

get eight big horizontal separated pictures instead of one screen.

radeon x1950 catalyst 10.2 (latest for that card) with the .exe wrapper.

if i remember right i got the same result in an older exe test with the space shooter and another ati card.

os: xp, dx9 latest, 15fps.

tried to recreate the scenario ->

opened jump and run example 9.

set text to fps.

set fullscreen mode off.

webgl is disabled in browsers.

r84 faster than r95

firefox 11: -16%    

opera 11.62: -53%

the other results were in firefox 8 and opera 10 or 11 yesterday.

this time tested on a 3rd pc just to be sure.

so it could be something with canvas2d in windowed mode -

in chrome this doesnt seem to exist, in safari i remember small fps

decrease too(havent tested in safari that much)

the difference in opera is really high, in firefox its not that high but

firefox canvas2d performance is generally the worst thats why every fps is important in firefox. (tested all on lower to up middleclass hardware -

crysis1 10-40fps in medium 800x... 1280x... res,doom3 40-80fps mid to high, diablo3 15-100fps low-mid)

hi, working on a small game:

started out on r84 stable release (dont want to risk stability if a game doesnt need a new feature /commercial approach)

this week i exported with r95 (and r98) because of some improvements

and the performance in canvas2d went down in firefox for about 33%

and opera 50%. im using fairly new browser versions and it happens on 2

different pcs. chrome doesnt seem to be affected (or the framerate is too high to recognise it because the fps were locked at 60)

dont know if this is a bug or has to do with some perfomance enhancements which are mentioned in the changelogs.

sidenote: dont want to focus on webgl because after testing it and research of current support in browsers and hardware i cant use it or even recommend it(imo) to reach more users.

any experiences from other construct2 devs ?

Ashley do you know about this or any solutions ?

helps alot, thanks. (if it helps my game has to be seen <img src="smileys/smiley2.gif" border="0" align="middle" />

thx, the rounding works.

not for parallax enabled layers, which seems logic as i found no way to round that one too.

and scaling should be 1.0