JVET-N0003 JVET AHG report: Test model software development (AHG3) [F. Bossen, X. Li, A. Norkin, K. Sührung]
This report summarizes the activities of the AhG3 on test model software development that has taken place between the 13th and 14th JVET meetings.
The software development continued on the GitLab server. VTM version 3.2 was tagged on Jan. 19., VTM 4.0 on Feb. 12. and VTM 4.0.1 on Feb. 26. VTM 4.1 is expected during the 14th JVET meeting.
For core experiments (CEs) the same development workflow was followed as for the last meeting.
VTM software development
Development was continued on the GitLab server, which allows participants to register accounts and use a distributed development workflow based on git.
The server is located at:
https://vcgit.hhi.fraunhofer.de
The registration and development workflow is documented at:
https://vcgit.hhi.fraunhofer.de/jvet/VVCSoftware_VTM/wikis/VVC-Software-Development-Workflow
The VTM software can be found at
https://vcgit.hhi.fraunhofer.de/jvet/VVCSoftware_VTM/
VTM 3.2 was tagged on Jan 19, 2019.
- Changes relate to VTM 3.1 included:
- Removed JVET-Lxxx macros
- Fix for CU results reuse with dual tree (~7% encoder run time reduction for AI)
- Fix for block statistics generation
After one release candidates, VTM 4.0 was tagged on Feb. 12, 2019.
Changes related to VTM 3.2 include:
- JVET-M0453: modified CABAC core engine
- JVET-M0090: change default C[b,r]QpOffsetDualTree to 0, change C[b,r]QpOffset to 1 for CTC
- JVET-M0421: Split-first signalling for partitioning
- JVET-M0464: UniMTS
- JVET-M0479: 18-bit Mv clip
- JVET-M0502: context for prediction mode flag
- JVET-M0407: CPR range extension
- JVET-M0173: moving gtr2 flag to first pass
- JVET-M0142: CCLM collocated- hroma
- JVET-M0068/M0171: MMVD cleanup
- JVET-M0064: CCLM Simplification
- JVET-M0444: SMVD
- JVET-M0446/M0888/M0905: VPDU constraint at picture boundary
- JVET-M0487: integer positions in extended region
- JVET-M0497: Fast DST-7/DCT-8 with dual implementation support
- JVET-M0170: shared merging list candidate
- JVET-M0409: ATMVP fix when IBC is on
- JVET-M0118/M0185/etc: on merge_triangle_flag redudant signalling when a CU uses MMVD or CIIP
- JVET-M0381 Test CE2.2.2.a: one context for Merge_subblock_idx (AffMergeIdx in VTM software)
- JVET-M0145: Affine MV clip
- JVET-M0512: TMVP storage reduction using floating-point representation
- JVET-M0228/M0166/M0477: Remove CPMV comparisons for construnted affine merge candidates
- JVET-M0265: MV rounding cleanup
- JVET-M0255: MMVD without Fractional Distances for SCC
- JVET-M0273 (change 1)/JVET-M0240/JVET-M0116 (method 1)/JVET-M0338(method 1)/JVET-M0204(method 2):Only use left neighbour for ATMVP offset derivation
- JVET-M0264: Harmonization between HMVP and GBi
- JVET-M0063: BDOF bitDepth bug fix
- JVET-M0328: Only keep the second weight group during TPM MC
- JVET-M0192: modifications - affine chroma MV is average of two luma sub-block MVs
- JVET-M0193: allow only the first pairwise candidate, remove the remaining five
- JVET-M0883: Using regular merge index signalling for triangle mode
- JVET-M0257: Scan only non zero-out regions of large TUs (width or height > 32)
- JVET-M0312/M0255: disable fractional MVD for UHD sequences
- JVET-M0823: encoder optimization for MMVD
- JVET-M0246: adaptive MV precision for affine inter mode
- JVET-M0470: Fixed transition point between GR and TU + EG(k) escape codes for coefficient
- JVET-M0839: increase number of SATD candidates in Affine Merge from 1 to 2
- JVET-M0238: Nearest neighbour instead of linear interpolation in PDPC
- JVET-M0281 test a: perform all AMVP rounding before pruning even if AMVR is off
- JVET-M0436: Reduce HMVP table size
- JVET-M0117: On MVP candidate list generation for AMVP
- JVET-M0247: encoder optimization of affine AMVR
- JVET-M0471: Long deblocking filters
- JVET-M0908: CIIP deblocking
- JVET-M0303: implicit MTS
- JVET-M0427: Picture reconstruction with mapping
- JVET-M0119: remove scaling of ts coefficients
- JVET-M0253: Hash-based motion search
- JVET-M0147: DMVR (Decoder-side Motion Vector Refinement)
- JVET-M0102: Intra Sub-Partitions (ISP)
- JVET-M0126: HMVP merge pruing
- JVET-M0483: IBC
- JVET-M0297: skipping high frequency coefficients in wide/high blocks
- JVET-M0140 Sub-block Transform (SBT) for inter blocks
- JVET-M0854: MMVD precision for UHD (encoder only)
- JVET-M0055: VTM transcoding capabilities/DebugCTU
- removal of SPSNext / SPS alignment with draft text
- CABAC states init retraining
- Configuration file for class F
- Various bug fixes and cleanups
- Software manual updates
VTM 4.0.1 was tagged on Feb. 26, 2019. Changes include:
- Bug fixes
- JVET-M0277: PCM and loop filter
- JVET-M0428: Deblocking optimization
- JVET-M0451: Interoperability syntax
- JVET-M0600: Rate control
- JVET-M0102: encoder speedup for ISP
- Macro to disable MMVD
It should be noted that version 4.0.1 addresses an issue where sequential and parallel encoding modes for the random access (RA) configuration were yielding different results.
VTM 4.1 is expected to be tagged during the 14th JVET meeting. Changes include:
- JVET-M0132: APS implementation
- JVET-M0685: QP parameter predictor derivation
- JVET-M0113/M0188 quantization groups based on area
- JVET-M0111: disable GBI signalling when WP is enabled
- JVET-M0864: 2D cache model
- JVET M0445 MCTS
- JVET-M0091: sub-CTU QPA for <=HD sequences, share mean-value calculations via computeAvg( )
- SPS/SPSNext cleanup
- Various bug fixes and cleanups
At the beginning of the 14th meeting, some high level syntax implementations are still pending:
- JVET-L0686 Spec text for the agreed starting point on slicing and tiling (may be superseded by JVET-M0853)
- JVET-M0101: On VVC HLS (merge request submitted)
- JVET-M0415: horizontal reference wrap-around signalling (merge request submitted)
- JVET-M0853: tile groups PPS Signalling implementation (merge request for partial implementation submitted)
Common test conditions performance
The following table shows VTM 4.0 performance over HM 16.20:
|
| All Intra Main10 |
|
| |
|
| Over HM-16.20 |
|
| |
Y | U | V | EncT | DecT | |
Class A1 | -23.74% | -39.31% | -36.64% | 1375% | 170% |
Class A2 | -22.96% | -25.04% | -19.85% | 2155% | 179% |
Class B | -19.12% | -24.27% | -31.03% | 2341% | 179% |
Class C | -19.52% | -21.31% | -24.54% | 3136% | 189% |
Class E | -22.80% | -23.91% | -27.15% | 1860% | 174% |
Overall | -21.23% | -26.19% | -28.01% | 2170% | 179% |
Class D | -16.30% | -18.23% | -19.63% | 3471% | 184% |
Class F | -33.58% | -36.53% | -38.27% | 3251% | 183% |
|
| Random Access Main10 |
|
| |
|
| Over HM-16.20 |
|
| |
Y | U | V | EncT | DecT | |
Class A1 | -31.64% | -43.55% | -45.48% | 717% | 174% |
Class A2 | -36.35% | -40.36% | -35.20% | 800% | 170% |
Class B | -30.96% | -42.07% | -42.78% | 780% | 161% |
Class C | -26.08% | -31.79% | -33.50% | 941% | 173% |
Class E |
|
|
|
| |
Overall | -30.87% | -39.28% | -39.33% | 810% | 169% |
Class D | -25.20% | -29.25% | -30.34% | 989% | 176% |
Class F | -35.75% | -40.34% | -41.42% | 498% | 137% |
|
| Low Delay B Main10 |
|
| |
|
| Over HM-16.20 |
|
| |
Y | U | V | EncT | DecT | |
Class A1 |
|
|
|
|
|
Class A2 |
|
|
|
| |
Class B | -24.33% | -33.19% | -32.22% | 745% | 150% |
Class C | -21.17% | -27.80% | -28.90% | 853% | 156% |
Class E | -24.36% | -30.57% | -33.01% | 382% | 119% |
Overall | -23.29% | -30.74% | -31.31% | 660% | 143% |
Class D | -21.07% | -24.72% | -25.38% | 843% | 159% |
Class F | -31.86% | -38.71% | -39.89% | 421% | 117% |
The following table shows VTM 4.0 performance compared to VTM 3.2:
|
| All Intra Main10 |
|
| |
|
| Over VTM-3.2 |
|
| |
Y | U | V | EncT | DecT | |
Class A1 | -2.66% | -3.82% | -4.59% | 132% | 96% |
Class A2 | -2.65% | -2.97% | -2.75% | 126% | 103% |
Class B | -2.18% | -1.75% | -2.17% | 124% | 104% |
Class C | -2.40% | 1.58% | 1.21% | 124% | 117% |
Class E | -2.45% | 1.57% | 1.27% | 130% | 111% |
Overall | -2.43% | -1.01% | -1.35% | 126% | 107% |
Class D | -2.18% | 0.09% | 0.08% | 117% | 115% |
Class F | -18.38% | -14.24% | -14.34% | 211% | 111% |
|
| Random Access Main10 |
|
| |
|
| Over VTM-3.2 |
|
| |
Y | U | V | EncT | DecT | |
Class A1 | -5.43% | 6.44% | 3.69% | 147% | 115% |
Class A2 | -6.83% | 5.50% | 6.18% | 156% | 119% |
Class B | -5.04% | 5.89% | 5.33% | 153% | 124% |
Class C | -5.16% | 7.10% | 7.61% | 157% | 127% |
Class E |
|
|
|
| |
Overall | -5.51% | 6.25% | 5.78% | 153% | 122% |
Class D | -4.90% | 6.63% | 6.93% | 170% | 126% |
Class F | -17.37% | -9.37% | -9.08% | 164% | 126% |
|
| Low Delay B Main10 |
|
| |
|
| Over VTM-3.2 |
|
| |
Y | U | V | EncT | DecT | |
Class A1 |
|
|
|
|
|
Class A2 |
|
|
|
| |
Class B | -4.15% | 19.69% | 20.48% | 151% | 115% |
Class C | -4.19% | 14.54% | 15.70% | 157% | 131% |
Class E | -1.95% | 9.93% | 12.19% | 142% | 124% |
Overall | -3.61% | 15.53% | 16.81% | 151% | 122% |
Class D | -4.48% | 15.65% | 18.69% | 168% | 121% |
Class F | -13.29% | 0.70% | 1.13% | 151% | 125% |
The following table shows VTM 4.0 performance compared to VTM 3.2 under old CTC (no chroma QP adjustment):
|
| All Intra Main10 |
|
| |
|
| Over VTM-3.2 |
|
| |
Y | U | V | EncT | DecT | |
Class A1 | -2.66% | -3.82% | -4.59% | 132% | 97% |
Class A2 | -2.66% | -3.00% | -2.77% | 126% | 102% |
Class B | -2.18% | -1.77% | -2.21% | 124% | 105% |
Class C | -2.40% | 1.58% | 1.21% | 124% | 117% |
Class E | -2.45% | 1.57% | 1.27% | 130% | 113% |
Overall | -2.43% | -1.02% | -1.36% | 126% | 107% |
Class D | -2.17% | 0.09% | 0.09% | 117% | 114% |
Class F | -18.38% | -14.24% | -14.33% | 211% | 112% |
|
| Random Access Main10 |
|
| |
|
| Over VTM-3.2 |
|
| |
Y | U | V | EncT | DecT | |
Class A1 | -4.07% | -2.33% | -3.32% | 148% | 115% |
Class A2 | -5.27% | -3.04% | -2.29% | 157% | 119% |
Class B | -4.54% | -3.22% | -2.95% | 154% | 121% |
Class C | -4.60% | 0.20% | 0.51% | 157% | 127% |
Class E |
|
|
|
| |
Overall | -4.61% | -2.09% | -1.97% | 154% | 121% |
Class D | -4.39% | 0.41% | 0.73% | 170% | 125% |
Class F | -16.87% | -12.89% | -12.71% | 165% | 127% |
|
| Low Delay B Main10 |
|
| |
|
| Over VTM-3.2 |
|
| |
Y | U | V | EncT | DecT | |
Class A1 |
|
|
|
|
|
Class A2 |
|
|
|
| |
Class B | -3.31% | -1.45% | -0.41% | 152% | 116% |
Class C | -3.20% | -0.79% | -0.20% | 156% | 129% |
Class E | -1.48% | -0.78% | 0.28% | 143% | 125% |
Overall | -2.82% | -1.06% | -0.17% | 151% | 122% |
Class D | -3.65% | -1.46% | 0.11% | 169% | 122% |
Class F | -12.22% | -8.78% | -8.63% | 151% | 123% |
The following table shows VTM 4.0.1 performance compared to VTM 4.0:
All Intra Main10 | |||||
Over VTM-4.0 | |||||
Y | U | V | EncT | DecT | |
Class A1 | 0.03% | -0.01% | 0.07% | 97% | 102% |
Class A2 | 0.03% | 0.11% | 0.02% | 97% | 101% |
Class B | 0.03% | -0.01% | 0.06% | 97% | 102% |
Class C | 0.04% | -0.03% | -0.03% | 96% | 102% |
Class E | 0.05% | 0.05% | 0.01% | 96% | 103% |
Overall | 0.04% | 0.02% | 0.03% | 97% | 102% |
Class D | 0.02% | -0.02% | -0.13% | 96% | 101% |
Class F | 0.07% | 0.08% | -0.01% | 97% | 103% |
|
| Random Access Main10 |
|
| |
|
| Over VTM-4.0 |
|
| |
Y | U | V | EncT | DecT | |
Class A1 | 0.02% | 0.00% | 0.00% | 101% | 100% |
Class A2 | -0.01% | -0.02% | -0.01% | 100% | 102% |
Class B | 0.00% | -0.10% | -0.15% | 101% | 100% |
Class C | -0.01% | -0.14% | -0.25% | 101% | 102% |
Class E |
|
|
|
| |
Overall | 0.00% | -0.08% | -0.12% | 101% | 101% |
Class D | 0.00% | -0.14% | 0.11% | 100% | 99% |
Class F | 0.21% | 0.00% | 0.05% | 101% | 102% |
Low Delay B Main10 | |||||
Over VTM-4.0 | |||||
Y | U | V | EncT | DecT | |
Class A1 | |||||
Class A2 | |||||
Class B | 0.00% | -0.40% | 0.16% | 100% | 101% |
Class C | -0.01% | 0.46% | 0.08% | 100% | 101% |
Class E | 0.05% | 1.75% | 0.47% | 101% | 101% |
Overall | 0.01% | 0.42% | 0.21% | 100% | 101% |
Class D | -0.02% | 0.06% | 0.38% | 100% | 102% |
Class F | 0.12% | -0.11% | -0.09% | 100% | 99% |
There is a notable performance drop for random access and low delay in class F. Bug fixes to address this issue have been submitted for inclusion in VTM 4.1.
Full results were attached to the AHG report as Excel files.
Multi-resolution streaming test conditions
An updated method is proposed in JVET-N0446. The contribution reports the following performance data. In the tables below, AS-RA and MR-RA denote test conditions typically used in adaptive streaming, when a source video may be downsampled prior to encoding at a lower bitrate if encoding downsampled video results in a better rate-distortion trade-off than encoding the original video at that bitrate. The choice of resolution-QP pairs in AS-RA is done once per sequence and is based on the anchor, with 4 resolution-QP points in total. The MR-RA corresponds to encoding the video at four spatial resolutions each at 4 QP (total of 16 points), then determining the convex hull for both anchor and test for the BD-rate calculations. The PSNR is calculated between the original source video and upsampled (if needed) reconstructed video. Both test conditions use closed-GOP encoding and 2s intra picture intervals.
BD-rate between VTM 4.0 and HM 16.19 using the JVET CTC, AS-RA and MR-RA test conditions for classes A1, A2 and B.
JVET RA CTC | AS-RA | MR-RA | ||||||||
Class | Seq. | Y [%] | U [%] | V [%] | Y [%] | U [%] | V [%] | Y [%] | U [%] | V [%] |
Class A1 | Tango2 | -34.57 | -49.75 | -47.42 | -29.18 | -43.36 | -37.71 | -28.85 | -48.96 | -43.85 |
FoodMarket4 | -34.05 | -32.94 | -35.92 | -31.05 | -31.39 | -30.98 | -30.14 | -35.07 | -35.31 | |
Campfire | -29.34 | -34.61 | -42.86 | -22.14 | -18.47 | -30.87 | -24.32 | -32.28 | -40.95 | |
Average | -32.65 | -39.10 | -42.07 | -27.46 | -31.07 | -33.19 | -27.77 | -38.77 | -40.03 | |
Class A2 | CatRobot1 | -40.26 | -46.15 | -39.99 | -40.38 | -45.56 | -40.64 | -39.68 | -45.03 | -39.94 |
DaylightRoad2 | -40.51 | -41.81 | -33.47 | -39.96 | -46.72 | -36.52 | -39.91 | -44.96 | -37.16 | |
ParkRunning3 | -31.53 | -18.89 | -16.40 | -25.96 | -22.82 | -19.69 | -23.30 | -19.85 | -17.41 | |
Average | -37.43 | -35.62 | -29.95 | -35.44 | -38.37 | -32.28 | -34.29 | -36.61 | -31.50 | |
Class B | MarketPlace | -32.06 | -33.91 | -40.09 | -29.28 | -27.71 | -28.04 | -28.81 | -32.15 | -33.57 |
RitualDance | -27.99 | -30.08 | -37.09 | -27.14 | -29.30 | -31.95 | -26.57 | -31.84 | -34.86 | |
Cactus | -32.93 | -39.15 | -31.87 | -36.29 | -44.90 | -41.35 | -32.50 | -39.60 | -34.96 | |
BasketballDrive | -32.50 | -41.11 | -41.53 | -35.29 | -43.09 | -39.45 | -29.14 | -40.47 | -37.55 | |
BQTerrace | -31.07 | -40.71 | -40.11 | -32.91 | -45.74 | -50.67 | -26.67 | -39.38 | -42.59 | |
Average | -31.31 | -36.99 | -38.14 | -32.18 | -38.15 | -38.29 | -28.74 | -36.69 | -36.71 | |
Overall Average | -33.35 | -37.19 | -36.98 | -31.78 | -36.28 | -35.26 | -29.99 | -37.24 | -36.19 | |
BD-rate between VTM 4.0 and VTM 3.0 using the JVET CTC, AS-RA and MR-RA test conditions for classes A1, A2 and B.
JVET RA CTC | AS-RA | MR-RA | ||||||||
Class | Seq. | Y [%] | U [%] | V [%] | Y [%] | U [%] | V [%] | Y [%] | U [%] | V [%] |
Class A1 | Tango2 | -5.95 | 8.82 | 6.48 | -4.79 | 6.25 | 4.11 | -4.86 | 4.23 | 3.59 |
FoodMarket4 | -4.22 | 4.38 | 3.99 | -4.66 | 4.94 | 6.08 | -4.64 | 3.10 | 2.87 | |
Campfire | -6.15 | 6.08 | 0.67 | -1.52 | 1.40 | 4.95 | -3.77 | 5.06 | -0.17 | |
Average | -5.44 | 6.43 | 3.71 | -3.66 | 4.20 | 5.05 | -4.42 | 4.13 | 2.10 | |
Class A2 | CatRobot1 | -5.23 | 3.96 | 5.11 | -6.12 | 1.24 | 1.71 | -6.12 | 1.36 | 2.79 |
DaylightRoad2 | -6.09 | 6.55 | 6.86 | -6.63 | 0.64 | 1.75 | -6.43 | 4.89 | 7.02 | |
ParkRunning3 | -9.24 | 6.16 | 6.74 | -6.06 | 2.47 | 2.64 | -6.66 | 4.87 | 5.49 | |
Average | -6.85 | 5.56 | 6.24 | -6.27 | 1.45 | 2.03 | -6.40 | 3.70 | 5.10 | |
Class B | MarketPlace | -7.33 | 3.58 | 1.94 | -3.71 | 16.74 | 16.56 | -1.78 | 10.99 | 8.51 |
RitualDance | -4.76 | 4.04 | 4.90 | -5.28 | 10.17 | 12.15 | -3.09 | 6.19 | 8.36 | |
Cactus | -3.70 | 4.85 | 4.42 | -3.92 | 1.52 | 1.91 | -3.68 | 1.94 | 2.63 | |
BasketballDrive | -4.90 | 6.93 | 7.57 | -5.79 | 8.70 | 10.47 | -3.69 | 9.93 | 9.79 | |
BQTerrace | -4.60 | 10.99 | 8.22 | -4.39 | 7.67 | 4.84 | -5.40 | 5.01 | 6.65 | |
Average | -5.06 | 6.08 | 5.41 | -4.62 | 8.96 | 9.18 | -3.53 | 6.81 | 7.19 | |
Overall Average | -5.65 | 6.03 | 5.17 | -4.81 | 5.61 | 6.11 | -4.56 | 5.23 | 5.23 | |
Differences in CTC between HM and VTM
A difference in coding conditions between HM and VTM was found as follows:
JCT-VC decided at some point that parameter sets should be repeated at every IRAP picture. With the joint call for proposals for VVC, this was not made a requirement. The VVC CTC were inherited from the CfP conditions. VTM did not have the ability to repeat parameter sets, while the HM CTC configuration files still used the parameter.
A configuration parameter for repetition of parameter sets was implemented on top of VTM 4.0 and included in VTM 4.0.1. The following table shows the performance impact of repeating parameter sets (AI and RA configurations):
All Intra Main10 |
|
| |||
Over VTM-4.0 |
|
| |||
Y | U | V | EncT | DecT | |
Class A1 | 0.06% | 0.06% | 0.06% | 100% | 101% |
Class A2 | 0.03% | 0.03% | 0.03% | 100% | 101% |
Class B | 0.10% | 0.09% | 0.10% | 100% | 100% |
Class C | 0.26% | 0.23% | 0.24% | 100% | 99% |
Class E | 0.34% | 0.30% | 0.31% | 100% | 99% |
Overall | 0.16% | 0.14% | 0.15% | 100% | 100% |
Class D | 0.76% | 0.67% | 0.69% | 100% | 97% |
Class F | 0.26% | 0.29% | 0.19% | 101% | 100% |
|
| Random Access Main10 |
|
| |
|
| Over VTM-4.0 |
|
| |
Y | U | V | EncT | DecT | |
Class A1 | 0.01% | 0.01% | 0.01% | 98% | 100% |
Class A2 | 0.01% | 0.01% | 0.01% | 98% | 99% |
Class B | 0.03% | 0.03% | 0.03% | 99% | 97% |
Class C | 0.08% | 0.07% | 0.07% | 99% | 100% |
Class E |
|
|
|
| |
Overall | 0.03% | 0.03% | 0.03% | 99% | 99% |
Class D | 0.22% | 0.21% | 0.21% | 99% | 97% |
Class F | 0.14% | 0.06% | -0.12% | 99% | 98% |
In the discussion, it was agreed to repeat the parameter sets at random access points in VTM testing.
Software manual
The software manual was updated to reflect general scope and compilation instructions. The encoding and decoding parameter description though still reflects HM, rather than VTM.
Although the software coordinators requested contributions to the software manual for changed parameters twice on the JVET email reflector, not a single input was received.
Update to Guidelines for VVC reference software development
A proposed update for the guidelines was uploaded as JVET-N0503.
CE software
For each CE a group was created in GitLab and CE coordinators were given owner rights to the group. This way they could clone VTM as required, create branches for different tests and assign user access to the group themselves.
The CE development workflow is described at:
https://vcgit.hhi.fraunhofer.de/jvet/VVCSoftware_VTM/wikis/Core-experiment-development-workflow
CE read access is available using shared accounts: One account exists for MPEG members, which uses the usual MPEG account data (as announced on the appropriate email lists). A second account exists for VCEG members. The account information for VCEG members is available in the TIES system:
https://www.itu.int/ifa/t/2017/sg16/exchange/wp3/q06/vceg_account.txt
Bug tracking
The bug tracker for VTM and specification text is located at:
https://jvet.hhi.fraunhofer.de/trac/vvc
The bug tracker uses the same accounts as the HM software bug tracker. Users may need to log in again due to the different sub-domain. For spam fighting reasons account registration is only possible at the HM software bug tracker at
https://hevc.hhi.fraunhofer.de/trac/hevc
Please file all issues related to the VVC reference software into the bug tracker. Try to provide all the details, which are necessary to reproduce the issue. Patches for solving issues and improving the software are always appreciated.
Recommendations
The AHG recommended to:
- Continue to develop the VTM reference software
- Improve documentation, especially the software manual
- Resolve any normative issues resulting from the large number of integrations in the most recent development cycle
- Encourage people to test VTM software more extensively outside of common test conditions.
- Encourage people to report all (potential) bugs that they are finding.
- Encourage people to submit bit-streams/test cases that trigger bugs in VTM.
- Encourage people to submit non-normative changes that reduce encoder run time without significantly sacrificing compression performance
- Review and approve updated guidelines for reference software development