Fixed in ECM-9.1
The ECM-9.0 encoder crashes when encoding with RPR (LDB configuration tested). Disabling the following two macros seems to fix the problem:
Jonatan Samuelsson-Allendes (ee756301) at 25 Jul 00:35
JVET-AE0150: RPR filters and thresholds for scale factors 1.1x to 1.35x
Jonatan Samuelsson-Allendes (ee756301) at 24 Jul 15:22
White space alignment
Jonatan Samuelsson-Allendes (2b9ef71b) at 24 Jul 15:20
JVET_AE0150_SMALL_SCALE_RPR_FILTERS
Jonatan Samuelsson-Allendes (dd713b31) at 24 Jul 14:33
Merge branch 'jvet-ae-0100-bvgcccm' into 'master'
... and 85 more commits
Two example command lines that exhibit the problem(s):
EncoderApp.exe -c cfg/encoder_lowdelay_ecm.cfg -i BQSquare_416x240_60.yuv -o out.yuv -wdt 416 -hgt 240 -fr 30 -q 37 -f 300 -c cfg/rpr/scale1.5x.cfg
crashes after POC 7 (at the first resolution switch) with the following error message:
ERROR: In function "InterPrediction::getAffAMLRefTemplateMvdPredUni" in C:\Users\jonat\proj\sharp_ecm_ee\source\Lib\CommonLib\InterPrediction.cpp:11997: getAffAMLRefTemplate not supported with ref scaled.
EncoderApp.exe -c encoder_lowdelay_ecm.cfg -i RaceHorses_416x240_30.yuv -o out.yuv -wdt 416 -hgt 240 -fr 30 -q 37 -f 300 --RPR=1 --ScalingRatioHor=1.25 --ScalingRatioVer=1.25 --FractionNumFrames=0.5 --UpscaledOutput=2
crashes after POC 1 (well before the first resolution switch is supposed to happen) with a Segmentation fault
It would be great if these problems with the RPR functionality in ECM-9.0 can be fixed as soon as possible since the RPR functionality is needed for the tests in EE2 (test 3.7a and 3.7b).
The ECM-9.0 encoder crashes when encoding with RPR (LDB configuration tested). Disabling the following two macros seems to fix the problem:
Jonatan Samuelsson-Allendes (c6b0d6a2) at 11 May 20:43
Updating comment to describe new filter
OK, sounds good.
I'm not sure what that comment is intended to say. Is it an indication of the scale factor the filter is optimized for? Or is it intended to describe some property of the filter (such as the stretch of the sinc function)? In either case the comments for the filters following this one seems incorrect; the filter with comment "D = 2" is typically used for 1.5x scaling and the filter with comment "D = 2.5" is typically used for 2x scaling. Should we perhaps change them to "D = 1.33", "D = 1.5" and "D = 2", respectively?
Or perhaps even add a bit of description of he filters such as:
// Kaiser(7)-windowed sinc ratio 1.35
// cos-windowed sinc ratio 1.5, 0.9π-cutoff
// cos-windowed sinc ratio 2, 0.9π-cutoff
Jonatan Samuelsson-Allendes (d71178fe) at 08 May 23:36
Jonatan Samuelsson-Allendes (cace117c) at 08 May 19:37
Fix typo in macro JVET_AD0169_SMALL_SCALE_DOWNSAMPLING
... and 18 more commits
Jonatan Samuelsson-Allendes (7752a627) at 08 May 19:34
Update version number to 20.2
... and 16 more commits