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

Re: [Public WebGL] Support precompiled shaders as extensions



Should be a good thing to add even without support on non-Windows
platforms.  It can always report just one thread and block when you kick
it off on systems that don't have parallel compilation.

  -- Steve

Florian Bösch wrote:
> Do you think that D3D would be capable of implementing
> ARB_parallel_shader_compile?
>
> I'm asking because ARB_parallel_shader_compile offers two new interfaces:
>
>    - glMaxShaderCompilerThreadsARB(count) -> sets the maximum number of
>    threads to use for shader compilation
>    - glGetShaderiv/glGetProgramiv (would be
>    gl.getShaderParameter/gl.getProgramParameter in WebGL) with the
> parameter
>    COMPLETION_STATUS_ARB -> asynchronously confirms if the shader is still
>    compiling or the program is still linking
>
> If D3D can support that functionality (maybe with some help from the
> browser), then this extension could be exposed to WebGL, and would not
> only
> allow parallel compilation per-se, but also to control how many threads
> are
> used for it, and to get at least some kind of asynchronous progress report
> about it.
>
> This functionality is so far not supported by OpenGL ES extensions or any
> version of ES core. Though I believe if the Khronos WebGL-WG asked the
> ES-WG if they could introduce it, chances would not be bad.
>
> On Fri, Nov 18, 2016 at 9:05 PM, Jamie Madill <jmadill@chromium.org>
> wrote:
>
>> RE slow shader compiles, one improvement we're planning to implement
>> very
>> soon is parallel compiles for the PS and VS on D3D in ANGLE. If the host
>> has the CPU for this it could speed things *up* to 2x (3x for point
>> sprite programs). This doesn't solve the problem, but should help.
>>
>> On Wed, Nov 16, 2016 at 8:44 AM, Florian Bösch <pyalot@gmail.com>
>> wrote:
>>
>>> On Wed, Nov 16, 2016 at 1:09 PM, Mr F <arthur@playcanvas.com> wrote:
>>>
>>>> Something like WebDX11 (but not limited to Win) would be perfectly
>>>> nice
>>>> :D
>>>>
>>>
>>> As I've explained, the problem is not the frontend syntax, but the
>>> backend. It's not an acceptable solution to solve something that can
>>> only
>>> work for a small (and shrinking) minority of the web. If you would find
>>> that (WebDX) perfect, we might as well do WebMetal, which only runs on
>>> iOS,
>>> would you find that perfect?
>>>
>>
>>
>


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