Said Abou-Hallawa

I have few thoughts about the two proposals in https://github.com/fserb/canvas2D/blob/master/spec/filters.md: For the main proposal: - It lacks a clear definition for the primitive inputs and result. It does not specify...

Thanks for fixing this after I broke it. The change looks good but I think it will be better if you get a separate applier for NEON. Instead of having...

> @saghul Yeah I'm also experiencing that blur doesn't work. `ctx.filter` is set to "none" by default instead of undefined so it seems like it's enabled. > > ```js >...

> Could this be tested? I am trying although it has been hard to mimic what lego.com is doing in a layout test.

> Great that you're back on this topic @shallawa - can you update us what changed? I just had a quick glance, and think you only rebased it, right? >...

> FWIW the switcher concept is confusing to me. The source of it seems to be that we hold the filter as a top level concept. The stack becomes: >...

> * SVG filter reference is resolved based on a RenderObject. This makes no sense, as it fails to resolve due to element not having a renderer. What it should...

To make this PR easier to review, I changed it to address the filter with the detached canvases only. So an `Element` and a `Document` are associated with the canvas....

Hi Johnny, Thanks for your review. I have not done extensive comparisons between the old and the new calculations. Our A/B system can't compare MotionMark changes. But I did compare...

Hi Johnny, The large confidence interval with the new rewrite happens because the optimal complexity is not clipped as before. This part in the original code removes the extreme results...