1339127 - Is there room to optimize the existing JS -> WebGL DOM bindings?

Status: NEW

(Core :: JavaScript Engine, defect, P3)

Description

[Jukka Jylänki]

Attachment: webgl_dom_validation_without_webgl_validation.png

Benchmarking some WebGL heavy applications, a significant bottleneck is caused by the DOM bindings when Wasm thunks to JS, and JS values are converted over to perform WebGL calls.

See the attached image for profiling data. A large proportion of time is spent in these bindings, marshalling values over.

a) Is there anything that shows up as potential for optimization via plain old shaving off clock cycles?

b) Luke mentioned that it might be possible to have a sufficiently smart JIT that would recognize these types of hot DOM calls, and it could optimize the WebGL bindings to avoid having to do costly type validation, but fast track the calls through. Would this be feasible to implement? Eventually we will have WebGL importable directly to Wasm in the distant future, though that optimization will be Wasm specific, but perhaps this kind of JIT optimization could be more generic in non-Wasm case as well?

Comment 1

[Jukka Jylänki]

(The eventual item of being able to call WebGL directly from Wasm is bug 935315)