diff options
| author | dalecurtis@chromium.org <dalecurtis@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98> | 2014-03-25 22:36:29 +0000 |
|---|---|---|
| committer | dalecurtis@chromium.org <dalecurtis@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98> | 2014-03-25 22:36:29 +0000 |
| commit | 9d85eea41488b044b467c7857f10e29bba9e3624 (patch) | |
| tree | d2a505fef105dec950384a1f8af5b6317440c547 /chrome/common/extensions/docs/server2/template_renderer.py | |
| parent | 715af36df18e2fe0d7ff7dd20eede32e6928db28 (diff) | |
| download | chromium_src-9d85eea41488b044b467c7857f10e29bba9e3624.zip chromium_src-9d85eea41488b044b467c7857f10e29bba9e3624.tar.gz chromium_src-9d85eea41488b044b467c7857f10e29bba9e3624.tar.bz2 | |
Fix over-looping in SincResampler. Reorder debug checks.
Several developers have complained about hitting the DCHECK on
virtual_source_idx_ < block_size_ with the previous for() loop
construct. Due to precision loss during repeated adding, the
previous code might iterate one more time than necessary.
The original for() loop was added to improve ARM performance by
avoiding repeated double < int comparisons. At the time I went
a bit farther and used a fixed calculation, which now appears
impractical. So I've reverted to the original int < int perf
improvement.
While digging around I also switched the debug checks for use
after free behavior to be more effective post-destruction (so
long as memory hasn't been clobbered).
Performance is unchanged.
BUG=none
TEST=SincResampler w/ 168 frame request size works.
Review URL: https://codereview.chromium.org/210033009
git-svn-id: svn://svn.chromium.org/chrome/trunk/src@259367 0039d316-1c4b-4281-b951-d872f2087c98
Diffstat (limited to 'chrome/common/extensions/docs/server2/template_renderer.py')
0 files changed, 0 insertions, 0 deletions