Bug 99703 – [chromium] Add additional memory management fields to WebGraphicsContext3D

• Status: RESOLVED
• Resolution: FIXED
• Component: New Bugs

[Christopher Cameron]

[chromium] Add additional memory management fields to WebGraphicsContext3D

[Christopher Cameron]

Created attachment 169378 [details]
Patch

[Christopher Cameron]

Comment on attachment 169378 [details]
Patch

Add new fields to WebGraphicsMemoryAllocation to allow for specifying
limits for when visible and not, and to allow for specifying priority
cutoffs (e.g, to allow specifying that backgrounded tabs should
retain only their visible textures).

Add sendManagedMemoryStatsCHROMIUM to WebGraphicsContext3D, and
add WebGraphicsManagedMemoryStats structure, to allow the GPU memory
manager to make more informed choices.

[WebKit Review Bot]

Please wait for approval from [e-mail], [e-mail], [e-mail], [e-mail] or [e-mail] before submitting, as this patch contains changes to the Chromium public API. See also https://trac.webkit.org/wiki/ChromiumWebKitAPI.

[Christopher Cameron]

Some additional context:

The old fields of the structure, gpuResourceSizeInBytes and suggestHaveBackbuffer are left in place so that this can roll in and not break anything.  When I've changed the Chromium side, I'll delete these variables.

I put a some extra things into these structures (in particular, the allocation counter and acks) in case they are helpful. If, when I have the whole thing wired up, they aren't helpful, I'll get rid of them (along with the deprecated fields). The strategy here is to avoid iterating in both WebKit and Chromium.

[Christopher Cameron]

Created attachment 169383 [details]
Patch

[Christopher Cameron]

Ping.

[James Robinson]

Comment on attachment 169383 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=169383&action=review

I really dislike adding things "just in case".  Let's figure out what we actually need and do that.  If we don't know whether we need something yet, we definitely do not.

> Source/Platform/chromium/public/WebGraphicsMemoryAllocation.h:92
> +struct WebGraphicsManagedMemoryStats {

WebKit API requires one header per type. see http://trac.webkit.org/wiki/ChromiumWebKitAPI

[Christopher Cameron]

Created attachment 170204 [details]
Patch

[Christopher Cameron]

I've put the structures in separate files.

I've removed the counter/ack from the patch.  I think it's fairly likely that these will be used in testing (if not in actual policy), but we can burn that bridge when we get there.  All other fields have concrete plans to be used.

[Christopher Cameron]

Ping.

[Christopher Cameron]

Ping.  I'd like to get this in before M24 branches.

[James Robinson]

Comment on attachment 170204 [details]
Patch

R=me

[WebKit Review Bot]

Comment on attachment 170204 [details]
Patch

Clearing flags on attachment: 170204

Committed r132686: <http://trac.webkit.org/changeset/132686>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.