Back to Search Document details
41st Meeting: by teleconference, CH, January 2026 2026-01-14 05:35
JVET AHG report: ECM software development (AHG6)
Abstract
This report summarizes the activities of the AHG6 on ECM software development that has taken place between the 40th and 41st JVET meetings.
JVET-AO0006 JVET AHG report: ECM software development (AHG6) [V. Seregin (chair), J. Chen, R. Chernyak, F. Le Léannec, K. Zhang (vice-chairs)]
  1. Software development

ECM software repository is located at https://vcgit.hhi.fraunhofer.de/ecm/ECM.

ECM software is based on VTM-10.0 with enabled MCTF including the update from JVET-V0056, and GOP32, which is very close to VTM-11.0.

VTM-11.0ecm anchor https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/tree/VTM11_ANC is used for ECM performance evaluation.

The following changes were integrated into ECM-19.0:

JVET-AN0168: Regression CCP fusion (Test 2.8a) [MR 962]

JVET-AN0101: Remove single model for ALF-CCCM (Test-5.3a) [MR 964]

JVET-AN0095: Quantization center shift (Test-4.2c*) [MR 965]

JVET-AN0121: split_cu_flag context change (Test 1.1g) [MR 968]

JVET-AN0098: Regularization for ALF-CCCM (Test 5.1a) [MR 967]

JVET-AN0093: Joint reordering GPM with affine and intra prediction (Test 3.4a) [MR 969]

JVET-AN0236: Generated merge candidates based on template cost (Test 3.1) [MR 966]

JVET-AN0203: Move RGPM to group 2 and add TM off implementation [MR 970]

JVET-AN0090: Adaptive subsampling filter selection for CCLM/intra-CCCM/ALF-CCCM (Test 2.6a) [MR 971]

JVET-AN0214: CABAC contexts retraining using ECM-18 data [MR 973]

JVET-AN0158: Longer tap interpolation filtering (Test 2.4) [MR 974]

Fixes:

Fix number of non zeroed-out coefficients to be encoded by RDOQ in LFNST/NSPT case [MR 960]

Collection of bugfixes [MR 961]

Bugfixes related to IBC and similar technologies [MR 963]

Fix JVET-AM0138 for 12 bit [MR 959]

Bugfix TIMD Intra merge [MR 958]

Fix decoder mismatch/crash in JVET_AM0138_ENHANCED_TMP_MERGE_LIST_TIMD_BV [MR 957]

Fixed CABAC bits statistics in ECM decoder [MR 956]

The following changes were integrated into VTM-11.0ecm anchor:

Use enum for split type [MR 955]

ECM-19.0 and VTM-11.0ecm19.0 were tagged on November 11, 2025.

CTC Performance

In this section, the ECM test results following the ECM SDR CTC configurations described in JVET-AI2017 are summarized.

ECM-19.0 performance over ECM-18.0 anchor is summarized in the tables below.

All Intra Main 10

Over ECM-18.0

Y

U

V

EncT

DecT

EncVmPeak

DecVmPeak

Class A1

-0.06%

0.23%

0.23%

100.2%

104.1%

100.4%

100.3%

Class A2

-0.09%

-0.23%

-0.20%

98.5%

103.2%

100.6%

100.4%

Class B

-0.06%

0.19%

0.37%

95.8%

96.5%

98.5%

100.2%

Class C

-0.13%

0.04%

0.02%

98.5%

102.9%

100.6%

100.1%

Class E

-0.13%

0.04%

0.17%

100.2%

103.1%

101.5%

100.0%

Overall

-0.09%

0.07%

0.14%

98.3%

101.4%

100.1%

100.2%

Class D

-0.13%

0.19%

0.16%

99.0%

101.2%

100.7%

100.0%

Class F

-0.29%

-1.20%

-1.49%

110.4%

117.8%

100.1%

100.1%

Class TGM

-0.01%

0.07%

0.06%

114.8%

104.4%

100.4%

100.1%

Random Access Main 10

Over ECM-18.0

Y

U

V

EncT

DecT

EncVmPeak

DecVmPeak

Class A1

-0.26%

-0.46%

-0.80%

102.5%

109.9%

100.2%

100.0%

Class A2

-0.24%

-0.90%

-0.68%

107.9%

112.3%

100.3%

100.2%

Class B

-0.19%

-0.57%

-0.87%

101.4%

102.1%

100.2%

100.3%

Class C

-0.27%

-0.84%

-0.68%

106.7%

110.3%

100.5%

100.0%

Class E

Overall

-0.24%

-0.68%

-0.77%

104.3%

107.8%

100.3%

100.2%

Class D

-0.24%

-0.33%

-0.62%

105.6%

108.0%

100.5%

100.0%

Class F

-0.11%

-0.91%

-1.49%

112.7%

117.7%

99.9%

99.9%

Class TGM

0.07%

-0.15%

-0.25%

119.6%

124.5%

100.1%

99.9%

Low delay B Main 10

Over ECM-18.0

Y

U

V

EncT

DecT

EncVmPeak

DecVmPeak

Class A1

Class A2

Class B

-0.10%

-0.85%

-0.60%

101.8%

104.8%

100.3%

99.6%

Class C

-0.29%

-1.27%

-1.06%

103.5%

104.5%

100.5%

100.2%

Class E

-0.11%

0.26%

0.24%

104.1%

99.2%

100.4%

99.6%

Overall

-0.17%

-0.71%

-0.54%

103.0%

103.3%

100.4%

99.8%

Class D

-0.29%

-1.19%

-1.82%

106.1%

108.0%

100.5%

100.1%

Class F

-0.14%

-1.02%

-1.69%

110.6%

117.7%

100.4%

99.9%

Class TGM

0.03%

-0.70%

-0.23%

112.5%

107.6%

96.5%

100.4%

Low delay P Main 10

Over ECM-18.0

Y

U

V

EncT

DecT

EncVmPeak

DecVmPeak

Class A1

Class A2

Class B

-0.02%

-1.16%

-0.54%

105.5%

112.4%

100.3%

100.1%

Class C

-0.10%

-0.63%

-0.77%

106.6%

104.2%

100.5%

100.1%

Class E

-0.02%

0.09%

0.40%

105.3%

104.3%

100.4%

100.0%

Overall

-0.05%

-0.67%

-0.39%

105.8%

107.6%

100.4%

100.1%

Class D

-0.15%

-2.66%

-2.10%

106.9%

112.3%

100.8%

100.1%

Class F

0.13%

-1.71%

-1.38%

109.4%

117.8%

99.9%

99.8%

Class TGM

0.05%

-0.50%

-0.15%

113.8%

113.4%

100.7%

100.0%

The below tables show ECM-19.0 performance comparing to VTM-11.0ecm19.0 anchor.

All Intra Main 10

Over VTM-11.0ecm19.0

Y

U

V

EncT

DecT

EncVmPeak

DecVmPeak

Class A1

-14.87%

-16.35%

-27.31%

1183.3%

578.9%

237.6%

326.7%

Class A2

-21.54%

-25.02%

-29.24%

1117.5%

618.4%

238.9%

326.5%

Class B

-15.11%

-23.02%

-20.90%

1050.8%

646.9%

178.3%

355.8%

Class C

-15.30%

-12.03%

-13.20%

1030.4%

649.6%

122.4%

394.7%

Class E

-19.69%

-23.36%

-21.40%

1005.3%

659.8%

141.4%

381.4%

Overall

-16.95%

-19.86%

-21.73%

1070.2%

633.0%

173.8%

358.0%

Class D

-13.04%

-8.61%

-9.49%

993.9%

692.2%

110.1%

414.3%

Class F

-31.03%

-35.19%

-35.37%

616.5%

656.3%

125.9%

374.4%

Class TGM

-43.73%

-49.29%

-48.58%

505.2%

645.6%

150.3%

351.7%

Random Access Main 10

Over VTM-11.0ecm19.0

Y

U

V

EncT

DecT

EncVmPeak

DecVmPeak

Class A1

-27.70%

-25.43%

-37.71%

1226.1%

1275.1%

153.4%

247.6%

Class A2

-30.98%

-35.54%

-40.94%

1181.5%

1462.9%

153.2%

247.5%

Class B

-25.57%

-34.66%

-31.70%

1037.5%

1254.2%

144.8%

268.3%

Class C

-27.32%

-24.94%

-25.55%

1155.1%

1555.8%

117.4%

321.1%

Class E

Overall

-27.55%

-30.40%

-33.11%

1133.0%

1374.4%

140.1%

272.5%

Class D

-28.00%

-24.79%

-26.22%

1084.2%

1651.9%

105.4%

360.4%

Class F

-33.69%

-37.31%

-38.21%

807.9%

865.5%

121.6%

296.0%

Class TGM

-43.03%

-48.96%

-48.99%

721.6%

745.4%

125.8%

280.9%

Low delay B Main 10

Over VTM-11.0ecm19.0

Y

U

V

EncT

DecT

EncVmPeak

DecVmPeak

Class A1

Class A2

Class B

-22.69%

-41.03%

-35.94%

1048.5%

1184.3%

167.3%

266.0%

Class C

-25.22%

-29.20%

-31.15%

1016.4%

1236.8%

125.8%

316.7%

Class E

-22.82%

-29.50%

-27.98%

1033.7%

732.0%

140.2%

295.6%

Overall

-23.57%

-34.21%

-32.35%

1034.0%

1065.4%

145.5%

289.4%

Class D

-26.68%

-30.77%

-32.61%

1040.5%

1416.4%

111.1%

352.8%

Class F

-31.47%

-41.72%

-42.29%

818.4%

855.4%

130.1%

294.8%

Class TGM

-41.32%

-51.82%

-51.52%

707.0%

643.7%

120.8%

282.6%

Low delay P Main 10

Over VTM-11.0ecm19.0

Y

U

V

EncT

DecT

EncVmPeak

DecVmPeak

Class A1

Class A2

Class B

-20.61%

-49.22%

-44.58%

922.2%

1169.2%

172.3%

293.6%

Class C

-23.22%

-37.72%

-39.26%

863.4%

1172.2%

127.1%

344.9%

Class E

-20.96%

-38.14%

-36.76%

935.2%

737.6%

144.1%

318.4%

Overall

-21.56%

-42.61%

-40.85%

905.3%

1042.9%

148.9%

316.1%

Class D

-25.94%

-41.54%

-42.37%

863.5%

1382.6%

111.6%

374.7%

Class F

-29.50%

-47.20%

-47.58%

765.2%

793.5%

130.5%

318.5%

Class TGM

-39.51%

-54.10%

-53.63%

739.6%

668.2%

126.8%

294.8%

  1. ECM memory consumption

ECM encoder memory consumption (VmPeak, GiB) is provided in ECM encoder log files and is summarized in the table below as maximum class-wise consumption rounded up to GiB.

AI

RA

LB

Class A (A1 & A2)

9

16

Class B

4

6

5

Class C

2

3

3

Class D

2

2

2

Class E

3

3

Class F

4

6

5

Class TGM

4

8

7

It is encouraged to care about memory allocation when developing and integrating tools into ECM. In particular, it is strongly recommended to re-use already existed memory wherever possible, rather than systematically allocating new memory, and allocate only the required amount of memory if new memory is needed.

Recommendations

The AHG recommended to:

  • Continue to develop ECM software.
  • Improve the software documentation.
  • Encourage people to report all (potential) bugs that they are finding using GitLab Issues functionality https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/issues.
  • Encourage people to submit merge requests fixing identified bugs.
  • Encourage people to continue working on ECM memory consumption reduction.
  • Encourage people to continue working on speeding up ECM encoder to reduce the simulation time.
Decisions
merged
The AHG recommended to: Continue to develop ECM software. Improve the software documentation. Encourage people to report all (potential) bugs that they are finding using GitLab Issues functionality https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/issues. Encourage people to submit merge requests fixing identified bugs. Encourage people to continue working on ECM memory consumption reduction. Encourage people to continue working on speeding up ECM encoder to reduce the simulation time.
Citation