diff options
author | hclam@chromium.org <hclam@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98> | 2011-02-14 18:13:00 +0000 |
---|---|---|
committer | hclam@chromium.org <hclam@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98> | 2011-02-14 18:13:00 +0000 |
commit | 6173e4b106bfa12b5ba954caf8daf2c465330532 (patch) | |
tree | 77b99857344771f990190e6ffca8803eaa3b3169 /chrome_frame/chrome_frame_helper_main.cc | |
parent | a332692bfdab2ce4383391886554e8ae9282805e (diff) | |
download | chromium_src-6173e4b106bfa12b5ba954caf8daf2c465330532.zip chromium_src-6173e4b106bfa12b5ba954caf8daf2c465330532.tar.gz chromium_src-6173e4b106bfa12b5ba954caf8daf2c465330532.tar.bz2 |
Revised SSE2 color conversion
I discovered there's average and shuffle instructions which simplify the
calculations of UV components. Also the extract instruction was very useful.
In terms of the algorithm I rearranged the memory reads to be sequential.
Overall this gives about 10% improvement on 64-bits machine.
On 32-bits machine this gives 20% improvement. If compiled as 32-bits code this is
quite slow compared to 64-bits, about 2x slow down. So working on MMX code will
be a direction for improvement.
BUG=None
TEST=I tried it, it looks right.
Review URL: http://codereview.chromium.org/6488014
git-svn-id: svn://svn.chromium.org/chrome/trunk/src@74827 0039d316-1c4b-4281-b951-d872f2087c98
Diffstat (limited to 'chrome_frame/chrome_frame_helper_main.cc')
0 files changed, 0 insertions, 0 deletions