[FFmpeg-devel] [PATCH] Add additional FFV1 fate tests

Peter B. pb at das-werkstatt.com
Wed Nov 6 22:37:55 CET 2013


On 11/05/2013 01:42 PM, Michael Niedermayer wrote:
>> --- /dev/null
>> +++ b/tests/ref/fate/ffv1-enc-v1-bff
>> @@ -0,0 +1,4 @@
>> +ca220f499d4bf37cc062ccb2e4a8b853 *tests/data/fate/ffv1-enc-v1-bff.avi
>> +1103420 tests/data/fate/ffv1-enc-v1-bff.avi
>> +5ae298fe228fc3a27a02812e6fbfff81 *tests/data/fate/ffv1-enc-v1-bff.out.rawvideo
>> +stddev: 7922.96 PSNR: 18.35 MAXDIFF:65296 bytes:  7603200/   608256
>> diff --git a/tests/ref/fate/ffv1-enc-v1-bgr0 b/tests/ref/fate/ffv1-enc-v1-bgr0
>> new file mode 100644
>> index 0000000..46f6b06
>> --- /dev/null
>> +++ b/tests/ref/fate/ffv1-enc-v1-bgr0
>> @@ -0,0 +1,4 @@
>> +75deb3f47c5d13da3607a0675f31d713 *tests/data/fate/ffv1-enc-v1-bgr0.avi
>> +615130 tests/data/fate/ffv1-enc-v1-bgr0.avi
>> +4d09cb2686347b39184c5e69acb80a09 *tests/data/fate/ffv1-enc-v1-bgr0.out.rawvideo
>> +stddev: 6821.10 PSNR: 19.65 MAXDIFF:65293 bytes:  7603200/   608256
> tiny_psnr is run here (and in some other cases) with the wrong
> parameters
> (tiny seems using 16bit) but the raw video is 8bit

Hm...
The principle is clear: 8bit tests for 8bit, 16bit tests for >8bit.

I've looked at "tests/fate-run.sh" and I'm confused by what you wrote:
You're listing encoding-tests, and say that tiny_psnr seems to be using
16bit.

When starting my ffv1.mak file, I copy/pasted the ffv1 entries from
tests/fate/vcodec.mak, and vsync1 is 8bit (yuv420p), and these tests
call tiny_psnr.
Therefore I'd assume that tiny_psnr would do 8bit - and >8bit being the
exception.

Furthermore, I've ran a grep on "tests/fate/*.mak", searching for
yuv422p >8bit and gbrp. It seems to me that all >8bit tests are decoding
a given file only. No encoding.


Is there any existing example calling ENCDEC for >8bits, so I could see
how it's handled there?

Thanks,
Pb


More information about the ffmpeg-devel mailing list