Port using actual CTU size instead of MAX_CU_SIZE at encoder from ecm/ECM!516
Test results with RA classB are as follows. The anchor is the current master tip (VTM-22-31d07037).
Over VTM-22-31d07037 | ||||||
---|---|---|---|---|---|---|
Y | U | V | EncT | DecT | VmPeak | |
Class A1 | #VALUE! | #VALUE! | #VALUE! | #DIV/0! | #DIV/0! | #DIV/0! |
Class A2 | #VALUE! | #VALUE! | #VALUE! | #DIV/0! | #DIV/0! | #DIV/0! |
Class B | 0.00% | 0.00% | 0.00% | 100.4% | 100.7% | 100% |
Class C | #VALUE! | #VALUE! | #VALUE! | #DIV/0! | #DIV/0! | #DIV/0! |
Class E | ||||||
Overall (Ref) | #VALUE! | #VALUE! | #VALUE! | #DIV/0! | #DIV/0! | #DIV/0! |
Class D | #VALUE! | #VALUE! | #VALUE! | #DIV/0! | #DIV/0! | #DIV/0! |
Class F | #VALUE! | #VALUE! | #VALUE! | #DIV/0! | #DIV/0! | #DIV/0! |
As the method only affects the case that non-MAX_CU_SIZE is used. Non-CTC setting (--CTUSize=64--MaxBTNonISlice=64) was tested below. 7% memory saving (VmPeak) is observed, which is about 200MB (average 2.96GB to average 2.76GB)
Over VTM-22-31d07037--CTUSize=64--MaxBTNonISlice=64 | ||||||
---|---|---|---|---|---|---|
Y | U | V | EncT | DecT | VmPeak | |
Class A1 | #VALUE! | #VALUE! | #VALUE! | #DIV/0! | #DIV/0! | #DIV/0! |
Class A2 | #VALUE! | #VALUE! | #VALUE! | #DIV/0! | #DIV/0! | #DIV/0! |
Class B | 0.00% | 0.00% | 0.00% | 100.3% | 101.2% | 93% |
Class C | #VALUE! | #VALUE! | #VALUE! | #DIV/0! | #DIV/0! | #DIV/0! |
Class E | ||||||
Overall (Ref) | #VALUE! | #VALUE! | #VALUE! | #DIV/0! | #DIV/0! | #DIV/0! |
Class D | #VALUE! | #VALUE! | #VALUE! | #DIV/0! | #DIV/0! | #DIV/0! |
Class F | #VALUE! | #VALUE! | #VALUE! | #DIV/0! | #DIV/0! | #DIV/0! |