Re: ScriptProcessorNode off the main thread (Was: Questioning the current direction of the Web Audio API)

2013-10-25 15:40, Robert O'Callahan skrev:
> On Fri, Oct 25, 2013 at 3:17 PM, Joseph Berkovitz < 
> <>> wrote:
>     Here are some of the issues with off-the-shelf workers that I see:
>     - Shared workers, while not having a per-node overhead problem,
>     seem too global in scope. It feels as though we want different
>     working storage for different nodes, even if they share the same
>     script. Global communication between nodes would probably lead to
>     bugs, not advantages.
>     - If there is one Dedicated worker per node, it's likely that
>     Dedicated workers will have to be created in great quantity since
>     Web Audio in many use cases is a very node-intensive system. The
>     setup overhead for dedicated workers may be too high.
>     - Dedicated workers imply postMessage-style communication via a
>     MessagePort, which suggests unnecessary communication overhead
>     since we don't need to allow passing of arbitrary data structures
>     between node workers and the rest of the environment. We can focus
>     on audio events and their results.
>     - Going forward, Web Audio workers do not even need to communicate
>     with arbitrary other pieces of the browser environment. They only
>     need to communicate with the machinery running the audio graph.
>     So I suggest we at least consider a new flavor of Worker that is
>     tailored for WebAudio, or else move to a distinct object altogether.
> The question is whether building an alternative off-main-thread JS 
> execution context could offer significant advantages over building on 
> DedicatedWorker, such as lower overhead if you have a lot of them. 
> Personally I don't see any reason to believe that there would be much 
> benefit.

I think that it would be good to get some real figures here. For 
instance, given the scenario of creating (& destroying) 100 workers per 
second (given a more-than-trivial music example, with more-than-trivial 
per-note graphs), with >100 workers alive concurrently, what would be 
the potential overhead compared to native nodes? (not sure if it's an 
extreme example - but it's certainly a probable use case)

Ideally, there shouldn't be much difference to native nodes, but in 
reality there would be a heavier set-up time and a memory usage 
overhead. But how much?

> Complex or slow Worker features simply shouldn't be used by processing 
> code. If they are used, it's unlikely to be any worse than having the 
> processing code go into an infinite loop, which we have to handle anyway.

Agree. I also have a feeling that we can handle the GC problem by 
evangelising & supporting GC-free programming practices (i.e. create 
objects & data in the setup-phase, but never create objects in the 
processing phase). E.g. the Java game programming community is used to 
this [1].



> Rob
> -- 
> Jtehsauts  tshaei dS,o n" Wohfy  Mdaon  yhoaus eanuttehrotraiitny  
> eovni le atrhtohu gthot sf oirng iyvoeu rs ihnesa.r"t sS?o  Whhei csha 
> iids  teoa stiheer :p atroa lsyazye,d  'mYaonu,r  "sGients  uapr,e  
> tfaokreg iyvoeunr, 'm aotr  atnod  sgaoy ,h o'mGee.t"  uTph eann dt 
> hwea lmka'n?  gBoutt  uIp  waanndt  wyeonut  thoo mken.o w *
> *

Marcus Geelnard
Technical Lead, Mobile Infrastructure
Opera Software

Received on Friday, 25 October 2013 15:59:37 UTC