Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(92)

Issue 2478313004: HalfFloat neon armv7 fix for destination pointer. (Closed)

Created:
4 years, 1 month ago by fbarchard1
Modified:
4 years, 1 month ago
Reviewers:
wangcheng, mthiffault
CC:
hubbe
Target Ref:
refs/heads/master
Project:
libyuv
Visibility:
Public.

Description

HalfFloat neon armv7 fix for destination pointer. Improved unittests detect different in arm64 rounding. TEST=util/android/test_runner.py gtest -s libyuv_unittest -t 7200 --verbose --release --gtest_filter=*Half* -a "--libyuv_width=640 --libyuv_height=360" BUG=libyuv:560 R=wangcheng@google.com Committed: https://chromium.googlesource.com/libyuv/libyuv/+/f2c27dafa2950510ba767cd59937ddf5d1974937

Patch Set 1 #

Patch Set 2 : add more tests for 16 bit #

Patch Set 3 : arm 32 bit C code does denormals #

Patch Set 4 : fix vmovl.u16 to expand shorts to ints #

Total comments: 2

Patch Set 5 : expect zero difference #

Patch Set 6 : removed is from comment #

Patch Set 7 : replace one with regular scale for 64 bit pending investigation into unittest failure #

Patch Set 8 : allow for rounding error on arm64 #

Patch Set 9 : LE 1 not EQ 1 for error #

Patch Set 10 : added caveat on rounding difference #

Unified diffs Side-by-side diffs Delta from patch set Stats (+70 lines, -27 lines) Patch
M README.chromium View 1 chunk +1 line, -1 line 0 comments Download
M include/libyuv/version.h View 1 chunk +1 line, -1 line 0 comments Download
M source/row_neon.cc View 1 2 3 2 chunks +8 lines, -8 lines 0 comments Download
M source/row_neon64.cc View 1 2 3 4 5 6 7 8 9 2 chunks +2 lines, -1 line 0 comments Download
M unit_test/planar_test.cc View 1 2 3 4 5 6 7 8 9 4 chunks +58 lines, -16 lines 0 comments Download

Messages

Total messages: 15 (5 generated)
fbarchard1
fix for ARMv7 HalfFloatPlane allow off by 1 on denormal test.
4 years, 1 month ago (2016-11-04 21:52:54 UTC) #2
fbarchard1
For the off by 1 the test just allows off by 1 for now, but ...
4 years, 1 month ago (2016-11-04 22:22:13 UTC) #3
fbarchard1
Fix in row_neon.cc for storing to correct location.
4 years, 1 month ago (2016-11-04 22:23:11 UTC) #6
fbarchard1
[ RUN ] LibYUVPlanarTest.TestHalfFloatPlane_16bit_denormal [ OK ] LibYUVPlanarTest.TestHalfFloatPlane_16bit_denormal (1 ms) [ RUN ] LibYUVPlanarTest.TestHalfFloatPlane_16bit_One [ ...
4 years, 1 month ago (2016-11-04 23:14:11 UTC) #7
wangcheng
https://codereview.chromium.org/2478313004/diff/50001/unit_test/planar_test.cc File unit_test/planar_test.cc (right): https://codereview.chromium.org/2478313004/diff/50001/unit_test/planar_test.cc#newcode2150 unit_test/planar_test.cc:2150: // 32 bit arm rounding on denormal is case ...
4 years, 1 month ago (2016-11-05 00:25:02 UTC) #8
wangcheng
4 years, 1 month ago (2016-11-05 00:25:05 UTC) #9
fbarchard1
https://codereview.chromium.org/2478313004/diff/50001/unit_test/planar_test.cc File unit_test/planar_test.cc (right): https://codereview.chromium.org/2478313004/diff/50001/unit_test/planar_test.cc#newcode2150 unit_test/planar_test.cc:2150: // 32 bit arm rounding on denormal is case ...
4 years, 1 month ago (2016-11-07 17:43:51 UTC) #10
fbarchard1
arm64 TestHalfFloatPlane_16bit_denormal (900 ms) TestHalfFloatPlane_16bit_One (806 ms) TestHalfFloatPlane_16bit_Opt (1040 ms) TestHalfFloatPlane_10bit_Opt (966 ms) TestHalfFloatPlane_9bit_Opt (936 ...
4 years, 1 month ago (2016-11-07 18:11:53 UTC) #11
wangcheng
lgtm
4 years, 1 month ago (2016-11-07 19:19:33 UTC) #12
fbarchard1
4 years, 1 month ago (2016-11-07 20:13:07 UTC) #15
Message was sent while issue was closed.
Committed patchset #10 (id:170001) manually as
f2c27dafa2950510ba767cd59937ddf5d1974937 (presubmit successful).

Powered by Google App Engine
This is Rietveld 408576698