WebGL Rendering Quality

[Jorgen Skogmo]

Im developing a WebGL application (using Three.js).

The attached image shows the project running on my MacBook Pro Retina 15", in

- Chrome 23.0.1271.101

- Safari 6.0.2 (8536.26.17)

- Node-webkit 0.3.7 (or 25.0.1324.0 as it's version string states...)

Im seeing pretty much the same performance,

(60 FPS when run alone, about 48 FPS when running two at the same time. Safari (always) perform poorly when other apps/windows are WebGL)

But: Quality is a lot worse in Node-webkit than any of the other.

Why is that?

Can it be improved?

Anything I can do?

I browsed the ejecta.js site yesterday, and saw them exposing an AA setting on the canvas element.

Not sure if its relevant to NW, but wanted to mention it anyway.

j.

[Roger WANG]

Thanks for reporting.

Could you please provide the code so I can see it here?

--
Roger WANG Intel Open Source Technology Center

[Roger WANG]

Jorgen Skogmo <j...@base.io> writes:

> Im developing a WebGL application (using Three.js).
> The attached image shows the project running on my MacBook Pro Retina 15",
> in
> - Chrome 23.0.1271.101
> - Safari 6.0.2 (8536.26.17)
> - Node-webkit 0.3.7 (or 25.0.1324.0 as it's version string states...)

btw, this issue from upstream seems to be your case:

http://code.google.com/p/chromium/issues/detail?id=159275

>
> Im seeing pretty much the same performance,
> (60 FPS when run alone, about 48 FPS when running two at the same time.
> Safari (always) perform poorly when other apps/windows are WebGL)
>
> But: Quality is a lot worse in Node-webkit than any of the other.
> Why is that?
> Can it be improved?
> Anything I can do?
>
> I browsed the ejecta.js site yesterday, and saw them exposing an AA setting
> on the canvas element.
> Not sure if its relevant to NW, but wanted to mention it anyway.
>
> j.
>

[Jorgen Skogmo]

Den mandag den 7. januar 2013 13.45.09 UTC+1 skrev Roger WANG:

Jorgen Skogmo <j...@base.io> writes:

> Im developing a WebGL application (using Three.js).
> The attached image shows the project running on my MacBook Pro Retina 15",
> in
> - Chrome 23.0.1271.101
> - Safari 6.0.2 (8536.26.17)
> - Node-webkit 0.3.7 (or 25.0.1324.0 as it's version string states...)

btw, this issue from upstream seems to be your case:

http://code.google.com/p/chromium/issues/detail?id=159275

Hi,

Sorry, can share the project yet - but have attached a NW project containing another Three.js example.

The upstream-issue concludes with blaming the multisampling in Apple's NVIDIA drivers...

Although that might be fully correct, I fail to understand how both current Chrome and Safari is able, then, to multisample...?!

Do you think this will be solved?

j.

[Roger WANG]

Jorgen Skogmo <j...@base.io> writes:

>> btw, this issue from upstream seems to be your case:
>>
>> http://code.google.com/p/chromium/issues/detail?id=159275
>>
>
> Hi,
>
> Sorry, can share the project yet - but have attached a NW project
> containing another Three.js example.

Thanks. Will check with it when I grab my Macbook.

> The upstream-issue concludes with blaming the multisampling in Apple's
> NVIDIA drivers...
> Although that might be fully correct, I fail to understand how both current
> Chrome and Safari is able, then, to multisample...?!

A wild guess: maybe the previous versions doesn't enable
multisampling... Anyway I'll check it. In the meantime, you could see
whether the command line arguments '--disable-gl-multisampling' works on
your side.

> Do you think this will be solved?

yes.

[Roger WANG]

Jorgen Skogmo <j...@base.io> writes:

>> btw, this issue from upstream seems to be your case:
>>
>> http://code.google.com/p/chromium/issues/detail?id=159275
>>
>
> Hi,
>
> Sorry, can share the project yet - but have attached a NW project
> containing another Three.js example.

The example failed to run:

"Uncaught ReferenceError: Detector is not defined", source:
file:///chromium/src/out/tests/webgl_lines_splines.nw/webgl_lines_splines.html (54)

>
> The upstream-issue concludes with blaming the multisampling in Apple's
> NVIDIA drivers...
> Although that might be fully correct, I fail to understand how both current
> Chrome and Safari is able, then, to multisample...?!
>
> Do you think this will be solved?

yes. probably by not using multisampling.

[Jorgen Skogmo]

The example failed to run:

"Uncaught ReferenceError: Detector is not defined", source:
file:///chromium/src/out/tests/webgl_lines_splines.nw/webgl_lines_splines.html (54)

yeah. Sorry about that... even though it does not complain, at all, here.

Find an fixed version attached.

 

> Do you think this will be solved?

yes. probably by not using multisampling.

Looking forward to see it :-)

 

[Thomas LAURENT]

Any news on that?