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

Re: [Public WebGL] More WebGL extensions




On Mar 31, 2011, at 12:33 PM, Kenneth Russell wrote:

> On Thu, Mar 31, 2011 at 12:20 PM, Chris Marrin <cmarrin@apple.com> wrote:
>> 
>> On Mar 31, 2011, at 12:00 PM, Kenneth Russell wrote:
>> 
>>>>> ...Also, we would need to specify interactions with OES_texture_float.
>>>>> Multisampled floating-point render targets would be extremely useful
>>>>> for high quality HDR rendering.
>>>> 
>>>> Sure, but that should be yet another extension, since it is possible to support one without the other.
>>> 
>>> I don't think a third extension is needed. All that's needed is to
>>> specify that if both the OES_texture_float and this multisampling
>>> extension are enabled, then multisampling is enabled for
>>> floating-point renderbuffers (if the implementation supports it -- and
>>> the framebuffer's status will be FRAMEBUFFER_UNSUPPORTED otherwise).
>> 
>> I don't know the reality in the drivers, but my concern is that there may be some drivers that support OES_texture_float and some form of framebuffer multisampling, but can't do multisampling with floating point texture buffers. The author needs to know such a situation exists to know how to allocate the supporting renderbuffers in the framebuffer.
>> 
>> But that may be a moot point. Maybe you always get one with the other?
> 
> Oops, my fault. I was confusing support for floating-point textures
> and floating-point renderbuffers. Multisampled FBOs require the use of
> renderbuffers, not textures, for the color attachment. Enabling
> OES_texture_float wouldn't affect the behavior of this multisampling
> extension.

Ah, right. The model would have to be render into a multisample FBO with a multisample renderbuffer then use a multisample blit (or glResolveMultisampleFramebufferAPPLE for iOS) to get the result into a normal, non-multisampled texture.
> 
> If we wanted that behavior, you're right, we'd need to spec another
> extension adding floating-point renderbuffers. I don't know whether
> there are any OES extensions for that functionality today.

Yeah, I haven't seen such a thing yet. But it's interesting that multisampling essentially results in higher precision pixel values. An interesting intermediate implementation could blit an integer multisample renderbuffer into a floating point texture. But that's probably not valuable enough for anyone to implement.

-----
~Chris
cmarrin@apple.com




-----------------------------------------------------------
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:
unsubscribe public_webgl
-----------------------------------------------------------