[YixinDu1]

At speed >= 4, pick the WIENER_NONSEP merge candidates and the frame filter class count from the accumulated filter statistics instead of filtering the frame once per candidate, and drop the largest loop restoration unit size one pyramid level earlier.

Anchor: commit 64c66e1
CTC: AV2 CTC v8 (33 frames, RA)

Speed 4 (cpu-used=4)
  +--------------+-------+-------+-------+-------+-------+--------+
  | Class        |     Y |    Cb |    Cr |  wAvg |  Enc% |  Ratio |
  +--------------+-------+-------+-------+-------+-------+--------+
  | A1           | -0.01 |  0.35 |  0.13 |  0.01 |  96.6 |  246.4 |
  | A2           |  0.01 |  0.18 |  0.08 |  0.02 |  99.0 |   50.0 |
  | Avg w/o B2   | -0.05 |  0.10 |  0.15 | -0.03 |  98.7 |  -39.2 |
  +--------------+-------+-------+-------+-------+-------+--------+

Speed 5 (cpu-used=5)
  +--------------+-------+-------+-------+-------+-------+--------+
  | Class        |     Y |    Cb |    Cr |  wAvg |  Enc% |  Ratio |
  +--------------+-------+-------+-------+-------+-------+--------+
  | A1           | -0.01 | -0.04 |  0.16 | -0.00 |  96.4 |    N/A |
  | A2           | -0.01 |  0.05 | -0.31 | -0.02 |  98.5 |  -84.3 |
  | Avg w/o B2   |  0.02 |  0.00 |  0.05 |  0.02 |  98.6 |   70.7 |
  +--------------+-------+-------+-------+-------+-------+--------+

Speed 6 (cpu-used=6)
  +--------------+-------+-------+-------+-------+-------+--------+
  | Class        |     Y |    Cb |    Cr |  wAvg |  Enc% |  Ratio |
  +--------------+-------+-------+-------+-------+-------+--------+
  | A1           | -0.03 | -0.02 |  0.14 | -0.02 |  95.3 | -217.6 |
  | A2           |  0.03 |  0.04 | -0.03 |  0.03 |  98.3 |   56.9 |
  | Avg w/o B2   |  0.00 | -0.04 | -0.27 | -0.01 |  98.1 | -208.8 |
  +--------------+-------+-------+-------+-------+-------+--------+

STATS_CHANGED

[urvangjoshi]

@YixinDu1 : looks like this PR was reopened?

[YixinDu1]

@YixinDu1 : looks like this PR was reopened?

Yes, we previously had to wait for our legal process to finish before reopen this PR

[YixinDu1]

Hi @kslu-aom, since you have also been working on LR, could you maybe help review this patch?

[kslu-aom]

Hi @kslu-aom, since you have also been working on LR, could you maybe help review this patch?

Only one comment above. This PR looks good to me otherwise.