[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Public WebGL] Odd behavior on eePC netbook.



Daniel Koch wrote:
>> ... this is a big fragment shader that hasn't been optimised yet -
>> and I'm surprised I didn't hit limits on my ancient nVidia 6800 box.
>> But Intel chipsets are by far the worst GPU's out there...so it's
>> entirely reasonable that I might have exceeded some limit.  It's also
>> possible that these are older drivers (although the EeePC netbook was
>> new out of it's box yesterday, there are no guarantees that the
>> manufacturer bothers to keep up with the latest driver releases).
>
> It is unfortunate, but we've certainly seen some of those come with
> pretty horrendous drivers (even for D3D).
>
With ANGLE running stand-alone, I get the same result.  Converting the
shader to HLSL and loading it up with a DirectX9 application produces no
useful errors either - so the underlying driver is simply being
non-communicative and there is evidently nothing that either ANGLE or
WebGL can do about that.

I hacked out a few features of the shader (I dropped normal mapping and
a couple of other 'eye candy' things that the game can do without) - and
it loaded up OK.  So I have things now where I have two versions of each
of the more complex shaders - one that I try to load first, the other as
a fallback if it fails for any reason whatever.

So I guess the advice for WebGL authors is that if you get any kind of a
failure on shader linkage at all - then try a simpler shader.

It's quite amazing to see what is still a relatively complex game
running on a $200 netbook inside a browser window!   I can't wait to get
this running on an iPhone or an Android!

Incidentally - I have located a WinXP machine with an even crappier
Intel chip...the 82865G...that fails to produce a WebGL rendering
context under either Firefox or Chrome...even though the Intel spec
sheet does claim DX9 support.  AFAICT, that chipset doesn't have shaders
of any kind - so unless they have a complete software fallback, it seems
unlikely that it's really doing DX9.

All efforts with loading up new drivers, upgrading DX9 and dropping to
16 bit color (all recommended by Intel as ways to get 3D games working)
have failed to improve on that situation.  So I guess there are still
some computers that are beyond help.

I wonder if it's worth starting a chipset compatibility chart on the
WebGL Wiki someplace?  It would certainly be useful for people who have
to support WebGL websites in the future to be able to point customers to
a "minimum specs" chart of some kind.

  -- Steve

-----------------------------------------------------------
You are currently subscribed to public_webgl@khronos.org.
To unsubscribe, send an email to majordomo@khronos.org with
the following command in the body of your email: