Bug 232840 – [GPU Process] [Filters 18/23] Remove the result FilterImage from FilterEffect

• Status: RESOLVED
• Resolution: FIXED
• Component: Layout and Rendering

[Said Abou-Hallawa]

A FilterEffect should be applied to any input FilterImages and results a FilterImage. These FilterImages will managed by the Filter class and should be cached by the creator of the source ImageBuffer. In the case of the GPU Process, they will be cached by RemoteResourceCache.

[Radar WebKit Bug Importer]

<rdar://problem/85425842>

[Said Abou-Hallawa]

Created attachment 448280 [details]
Patch

[Said Abou-Hallawa]

Created attachment 448302 [details]
Patch

[Darin Adler]

Comment on attachment 448302 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=448302&action=review

> Source/WebCore/platform/graphics/filters/FilterResults.cpp:40
> +    for (auto input : inputs) {

Using "auto" here means we copy each Ref<FilterImage>. I think that instead we should use "auto&" and save a bit of reference count churn.

> Source/WebCore/platform/graphics/filters/FilterResults.cpp:46
> +        auto addResult = m_resultReferences.ensure(input, [] {
> +            return FilterEffectSet();
> +        });
> +
> +        auto& references = addResult.iterator->value;
> +        references.add(effect);

This can just use add, because FilterEffectSet() is a null pointer, very cheap to construct and destruct. Can write this as a one-liner.

    m_resultReferences.add(input, { }).iterator->value.add(effect);

> Source/WebCore/svg/graphics/filters/SVGFilterBuilder.h:52
> +    void appendEffectToEffectRenderer(FilterEffect*, RenderObject*);

This should take a FilterEffect& instead of a FilterEffect*.

[Said Abou-Hallawa]

Created attachment 448605 [details]
Patch

[EWS]

Committed r287782 (245842@main): <https://commits.webkit.org/245842@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 448605 [details].