Bug 103797 – [chromium] Minimize memcpy for encoded image data in ImageFrameGenerator

• Status: RESOLVED
• Resolution: FIXED
• Component: Images

[Hin-Chung Lam]

The logic in ImageFrameGenerator regarding encoded image data is as follows:

1. When new data is received on the main thread, DeferredImageDecoder sets new data to ImageFrameGenerator.
2. ImageFrameGenerator makes a new copy of the data and keeps the reference.
3. When there is an image decoding request from LazyDecodingPixelRef, ImageFrameGenerator makes another copy of the data and take it for image decoding.

The entire process has too many unnecessary memcpys, it should be cleaned up.

[Hin-Chung Lam]

Created attachment 178331 [details]
Patch

[Hin-Chung Lam]

senorblanco, jamesr: Comments? This is the last big-ish change that is required for deferred image decoding. After this it will on par with existing main thread decoding stack.

[Stephen White]

Comment on attachment 178331 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=178331&action=review

Looks good.  r=me

> Source/WebCore/platform/graphics/chromium/ThreadSafeDataTransport.h:39
> +// The purpose of this class is to allow transferring data stored in

Grammar nit:  "transferring" -> "the transfer of"

> Source/WebCore/platform/graphics/chromium/ThreadSafeDataTransport.h:41
> +// SharedBuffer in a thread-safe manner. At the same time minimizing
> +// memory copy and thread contention.

Grammar nit:  this is not a sentence.  I'd just join it to the previous one:  ".. thread-safe manner, and to minimize memory copies and thread contention."

[Hin-Chung Lam]

Created attachment 179090 [details]
Patch for landing

[WebKit Review Bot]

Comment on attachment 179090 [details]
Patch for landing

Clearing flags on attachment: 179090

Committed r137508: <http://trac.webkit.org/changeset/137508>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.