JVET-AF0291 EE2 related: Encoder time reduction of Test-2.11 [K. Naser, F. Le Léannec, P. Bordes, A. Robert (InterDigital), S. Blasi, I. Zupancic, J. Lainema (Nokia), C. Zhou, Z. Lv (Vivo)] [late]
This contribution proposes an encoder-side optimization of tests in EE2-2.11 tests. It was asserted that this corresponds to a minimal change to the EE2-2.11x branch. The simulation results on AI for Test-2.11d with the encoder time reduction in this contribution are:
{ -0.12%, 0.02%, -0.06%, 100.9% (EncT) 98.6% (DecT) 100.3% (VmPeak)}
EE2 Test-2.11d was not adopted for two reasons: 1) the encoder runtime increase for AI was 4%, and 2) Test-2.11d is a combination of four tools.
The combined gain of 4 tools are now only 0.12% for AI, which means the individual gain for each tool is probably only ~0.03%.
Further study was recommended.
ECM modifications and software improvements beyond EE2 (41)
Contributions in this area were discussed at 1615–1910 on Saturday 14 Oct. 2023, 0840–1830 on Sunday 15 Oct. 2023 (chaired by Y. Ye), and on Monday 16 Oct. at 1715–1755 (chaired by JRO)