Re: [w3ctag/design-reviews] Incubation: CPU Performance API (Issue #1198)

RByers left a comment (w3ctag/design-reviews#1198)

We're discussing the [blink I2S](https://groups.google.com/a/chromium.org/g/blink-dev/c/igPwzkxhQtk) for this. For what it's worth I am convinced by the need for a static API (in addition to compute pressure) as it solves a different problem. For example, how should a video player decide whether to enable optional extra UI features (like the dynamic glow around youtube videos) while minimizing the risk of overloading the CPU and causing jank? Perhaps the most obvious case are games. AFAIK it's entirely standard practice for a game to configure default aspects of it's rendering settings based on the hardware capabilities (eg. draw distance or AI complexity) rather than try to predict and adapt dynamically. 

I don't understand why we wouldn't want to default to trusting the developers of advanced web apps like from Adobe and Figma who have said they need this. What such rich app developers really WANT (and get on other platforms) is full hardware details for every aspect of the device so they can be empowered to create the best possible user experience. Instead all we're talking about here is giving them *a couple bits* of information about the power for the CPU (which can be largely inferred if you're willing to waste a lot of CPU cycles). This seems like an incredibly pragmatic and webby tradeoff to me. Is the TAG taking the position that it knows better than these developers about the value of trying to solve such a problem?

> one thing we're skeptical about is the idea of leaving the buckets implementation-defined. Can you give us a rough algorithm that might be interoperable?

This is a GREAT question. I am also very [unsure about this](https://github.com/WICG/cpu-performance/issues/2). Does the TAG have any guidance on where on the spectrum it's better to fall between leaving such heuristics implementation defined, including a [full reference implementation](https://github.com/WICG/cpu-performance/pull/12) ([demo](http://rbyers.github.io/cpu-performance/demo.html)) normatively or non-normatively, or [putting some text in the spec](https://github.com/WICG/cpu-performance/pull/13/changes#diff-5e793325cd2bfc452e268a4aa2f02b4024dd9584bd1db3c2595f61f1ecf7b985) to cover most of the algorithm while leaving heuristics tweaks to impl-defined?


-- 
Reply to this email directly or view it on GitHub:
https://github.com/w3ctag/design-reviews/issues/1198#issuecomment-4760155385
You are receiving this because you are subscribed to this thread.

Message ID: <w3ctag/design-reviews/issues/1198/4760155385@github.com>