Closed GoogleCodeExporter closed 8 years ago
tried aug 18 build
The encoder thinks it worked
Pass 2/2 frame 150/150 3208B 171b/f 5132b/s 2104582 us (71.27 fps)
Stream 0 PSNR (Overall/Avg/Y/U/V) 100.000 100.000 100.000 100.000 100.000
But vpxdec repros the issue as well:
vpxdec --codec=vp9 --limit=10000 --i420 graytone0.vp9.webm -o
graytone_vp9.webm.854x480_30Hz_P420.yuv
psnr -s 854 480 graytone_vp9.webm.854x480_30Hz_P420.yuv
graytone.854x480_30Hz_P420.yuv
Frame PSNR-Y PSNR-U PSNR-V PSNR-All Frame
Global: 3.987018 128.000000 128.000000 5.747931 150
Avg: 3.987018 128.000000 128.000000 5.747931 150
Min: 3.987018 128.000000 128.000000 5.747931 0
Original comment by fbarch...@google.com
on 23 Aug 2013 at 6:30
1 pass also fails
vpxenc/dec were built like this:
CROSS=x86_64-w64-mingw32- ../libvpx/configure --target=x86_64-win64-gcc
--enable-static-msvcrt --disable-install-docs --disable-unit-tests
--disable-docs
make -i V=1
Original comment by fbarch...@google.com
on 23 Aug 2013 at 6:42
last known good version
vpxenc_jun14.exe VP9 Encoder v1.2.0-3097-g0b7910b
first known broken version
vpxenc_jul27.exe VP9 Encoder v1.2.0-3772-g118ccdc
Original comment by fbarch...@google.com
on 23 Aug 2013 at 7:42
Seems this is readily reproducible at speed 0/1/2 on linux as well:
v1.2.0-4091-g1c159c4.
After some bisecting the ssse3 version of vp9_quantize_b_32x32 is at least
partially to blame.
Original comment by jz...@google.com
on 24 Aug 2013 at 1:34
Looks like this is fixed, thanks!
vpxdec --codec=vp9 --limit=10000 --i420 graytone0.vp9.webm -o
graytone_vp9.webm.854x480_30Hz_P420.yuv
psnr -s 854 480 graytone_vp9.webm.854x480_30Hz_P420.yuv
graytone.854x480_30Hz_P420.yuv
Frame PSNR-Y PSNR-U PSNR-V PSNR-All Frame
Global: 78.714115 128.000000 128.000000 80.475027 150
Avg: 90.252552 128.000000 128.000000 91.567367 150
Min: 77.445382 128.000000 128.000000 79.206295 0
Original comment by fbarch...@google.com
on 30 Aug 2013 at 12:24
https://gerrit.chromium.org/gerrit/#/c/67004/
Original comment by jingn...@google.com
on 30 Aug 2013 at 1:59
Original issue reported on code.google.com by
fbarch...@google.com
on 23 Aug 2013 at 5:37