ector: move RLE shape/stroke computation to a pool of thread.
authorCedric Bail <cedric@osg.samsung.com>
Sun, 17 Sep 2017 06:23:36 +0000 (23:23 -0700)
committerCedric Bail <cedric@osg.samsung.com>
Sun, 17 Sep 2017 18:49:48 +0000 (11:49 -0700)
commite380ddb742384546ad05a5f8f5472ecdce795538
tree58cd8ea6ccf9ae8d170f1f6aa8995f085f29ef25
parent560e5c8e0f071d2cbaec0649bc1e8100d9e7f82e
ector: move RLE shape/stroke computation to a pool of thread.

This has been a long standing plan for improving performance in rendering
vector object. Depending on the test, you will get an improvement between
10 to 35% when rendering vector based object.

We are still maintaining the Cairo backend as the default one at the moment
due to a lack of result comparison tests between the two engine. Hopefully
we should get that covered and we can all enjoy a backend that is 4 times
faster by default.
src/lib/ector/software/ector_renderer_software_shape.c
src/lib/ector/software/ector_software_private.h
src/lib/ector/software/ector_software_rasterizer.c
src/lib/ector/software/ector_software_surface.c