Our performance gtests indicate anywhere from 10-20% reduction in
execution time based on the SSE2 version. Where it fell in the range
depended on the platform, but presumably that is related to the hardware
selected by treeherder. llvm-mca suggested it should be closer to 20%
on modern hardware (skylake).
Differential Revision: https://phabricator.services.mozilla.com/D55642
--HG--
extra : moz-landing-system : lando
Our performance gtests indicate anywhere from 10-20% reduction in
execution time based on the SSE2 version. Where it fell in the range
depended on the platform, but presumably that is related to the hardware
selected by treeherder. llvm-mca suggested it should be closer to 20%
on modern hardware (skylake).
Differential Revision: https://phabricator.services.mozilla.com/D55642
--HG--
extra : moz-landing-system : lando
No functional change. This allows us to support both alpha and non-alpha
variants with the same implementation, in addition to laying the ground
work for QCMS to support BGRA as an input and output.
Differential Revision: https://phabricator.services.mozilla.com/D30820
No functional change. This allows us to support both alpha and non-alpha
variants with the same implementation, in addition to laying the ground
work for QCMS to support BGRA as an input and output.
Differential Revision: https://phabricator.services.mozilla.com/D30820
Andrew Osmond
Csoregi Natalia
Narcis Beleuzu