GitLab

Currently when using OneCopyRasterBufferProvider we perform an upload
from CPU shared memory to the GPU and then a GPU to GPU copy during
raster playback. This CL allows direct upload of CPU shared memory
to the output shared image texture through the use of a new hint in
RasterDecoder::CopySubTexture that indicates CPU memory should be used
directly. This also allows us to unify how the copy is implemented
in RasterDecoder to always use Skia.

Bug: 984045
Change-Id: I6ac7d238cf3b2f43562481e2e1b36b3c4670a6f5
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2517149
Commit-Queue: Nathan Zabriskie <nazabris@microsoft.com>
Reviewed-by: Vasiliy Telezhnikov <vasilyt@chromium.org>
Reviewed-by: Sunny Sachanandani <sunnyps@chromium.org>
Cr-Commit-Position: refs/heads/master@{#841147}