Always have SATD based pruning in xCheckRDCostMerge2Nx2N
It is observed that under CTC (CIIP is enabled), always having SATD based pruning in xCheckRDCostMerge2Nx2N shows a bit better tradeoff. Moreover, it is easier for further refactoring.
The results relative to VTM-19 anchor are as follows.
| Random access Main10 | |||||
|---|---|---|---|---|---|
| Over vtm-19 | |||||
| Y | U | V | EncT | DecT | |
| Class A1 | -0.01% | -0.03% | 0.02% | 98% | 97% |
| Class A2 | -0.02% | -0.07% | 0.04% | 98% | 97% |
| Class B | 0.01% | 0.01% | -0.02% | 99% | 99% |
| Class C | -0.01% | 0.08% | 0.03% | 100% | 100% |
| Class E | |||||
| Overall | 0.00% | 0.00% | 0.01% | 99% | 98% |
| Class D | -0.06% | 0.04% | 0.13% | 100% | 99% |
| Class F | 0.00% | -0.01% | -0.02% | 100% | 99% |
| Low delay B Main10 | |||||
| Over vtm-19 | |||||
| Y | U | V | EncT | DecT | |
| Class A1 | |||||
| Class A2 | |||||
| Class B | -0.01% | 0.13% | -0.09% | 97% | 94% |
| Class C | 0.00% | 0.06% | 0.01% | 99% | 96% |
| Class E | 0.06% | -0.17% | 0.38% | 99% | 96% |
| Overall | 0.01% | 0.03% | 0.06% | 98% | 95% |
| Class D | 0.01% | -0.18% | 0.07% | 100% | 102% |
| Class F | 0.09% | 0.12% | -0.47% | 99% | 99% |
| Low delay P Main10 | |||||
| Over vtm-19 | |||||
| Y | U | V | EncT | DecT | |
| Class A1 | |||||
| Class A2 | |||||
| Class B | 0.01% | -0.05% | -0.01% | 97% | 94% |
| Class C | -0.01% | -0.31% | -0.09% | 99% | 95% |
| Class E | -0.10% | 0.44% | -0.04% | 98% | 95% |
| Overall | -0.02% | -0.01% | -0.04% | 98% | 95% |
| Class D | 0.05% | -0.75% | 0.28% | 100% | 102% |
| Class F | 0.04% | -0.03% | -0.50% | 99% | 96% |